Abstract
BACKGROUND: During the COVID-19 pandemic, a need for innovative, inexpensive, and simple ventilator devices for mass use has emerged. The Oxylator (CPR Medical Devices, Markham, Ontario, Canada) is an FDA-approved, fist-size, portable ventilation device developed for out-of-hospital emergency ventilation. It has not been tested in conditions of severe lung injury or with added PEEP. We aimed to assess the performance and reliability of the device in simulated and experimental conditions of severe lung injury, and to derive monitoring methods to allow the delivery of safe, individualized ventilation during situations of surge.
METHODS: We bench-tested the functioning of the device with an added PEEP valve extensively, mimicking adult patients with various respiratory mechanics during controlled ventilation, spontaneous breathing, and prolonged unstable conditions where mechanics or breathing effort was changed at every breath. The device was further tested on a porcine model (4 animals) after inducing lung injury, and these results were compared with conventional ventilation modes.
RESULTS: The device was stable and predictable, delivering a constant flow (30 L/min) and cycling automatically at the inspiratory pressure set (minimum of 20 cm H2O) above auto-PEEP. Changes in respiratory mechanics manifested as changes in respiratory timing, allowing prediction of tidal volumes from breathing frequency. Simulating lung injury resulted in relatively low tidal volumes (330 mL with compliance of 20 mL/cm H2O). In the porcine model, arterial oxygenation, CO2, and pH were comparable to conventional modes of ventilation.
CONCLUSIONS: The Oxylator is a simple device that delivered stable ventilation with tidal volumes within a clinically acceptable range in bench and porcine lung models with low compliance. External monitoring of respiratory timing is advisable, allowing tidal volume estimation and recognition of changes in respiratory mechanics. The device can be an efficient, low-cost, and practical rescue solution for providing short-term ventilatory support as a temporary bridge, but it requires a caregiver at the bedside.
Introduction
The coronavirus (COVID-19) pandemic has highlighted challenges with access to and possible critical shortages of medical equipment in settings with overwhelming needs.1-3 Unexpected surges of patients with life-threatening respiratory illnesses motivated some centers to consider splitting ventilators between patients,4,5 ventilating patients outside of the ICU, and using ventilators not designed for ICUs. These approaches have major limitations, and the capabilities and potential complications of these unconventional methods are poorly understood. Substantial government funding has led to scaled-up production of ventilators and a number of initiatives to design and build new, “simple” ventilators suitable for mass production in emergencies.6 Effective, inexpensive, and simple-to-use mechanical ventilators are needed to ventilate patients with COVID-19 at the onset of respiratory failure, in emergency rooms, outside of ICU settings, or during ICU stay and the prolonged recovery period.1-3,7,8 Such equipment may also be needed in future regional epidemic episodes or in under-resourced settings.
The Oxylator (CPR Medical Devices, Markham, Ontario, Canada), which has the Conformitè Europëenne mark and is FDA approved, is a fist-size, ultra-light, portable ventilation device developed for out-of-hospital use during resuscitation from cardiac arrest and emergency ventilation.9,10 It delivers patient-triggered or automated constant-flow, pressure-cycled breaths. It can be used with a PEEP valve placed in series and requires minimal adjustments (adjustable set pressure above PEEP). Limitations include a complete lack of monitoring displays, and settings are restricted to the pressure limit. Although the device is used in different countries during prehospital emergencies, it is unclear whether its use could be extended to situations of severe lung injury, especially given its lack of monitoring. It has never been tested in conditions of severe lung injury, such as that seen in COVID-19-related ARDS, nor in an ICU setting. Provided that the ventilation delivered can be predicted, it could be a useful device in critical situations as a bridge treatment until an ICU ventilator is available, or in resource-deprived settings where mechanical ventilators are unavailable or impractical. We aimed to assess the ventilation delivered and its predictability in terms of volume and rate based on pressure settings and respiratory mechanics, especially in simulated and experimental conditions of severe lung injury; and to derive bedside monitoring methods to allow the delivery of safe, individualized ventilation in situations of surge, such as during the COVID-19 pandemic.
QUICK LOOK
Current knowledge
During the COVID-19 pandemic, a need for innovative, inexpensive, and simple ventilator devices for mass use has emerged. The Oxylator is an FDA-approved, fist-size, portable ventilation device developed for out-of-hospital emergency ventilation. However, it has never been tested in conditions of severe lung injury or with an added PEEP valve.
What this paper contributes to our knowledge
In bench and porcine lung injury models, the Oxylator delivered stable ventilation with tidal volumes within a clinically acceptable range in lungs with low compliance. External monitoring of respiratory timing is advisable to allow bedside estimation of tidal volume and recognition of changes in respiratory mechanics. The device can be an efficient, low-cost, and practical rescue solution for providing short-term ventilatory support as a temporary bridge, but its use requires a caregiver to always be present.
Methods
We studied 6 individual Oxylator devices (4 EMX models, 2 HD models; weight = 0.25 kg) in a bench simulation and an animal model of lung injury. Experiments were primarily conducted with the EMX model, which is most commonly used worldwide. The bench study was performed at St. Michael’s Hospital, Toronto, Canada. Animal experiments were performed at The Hospital for Sick Children, Toronto, Canada.
Working Principle
The device (Fig. 1) can be connected to a tracheal tube, face mask, or supraglottic airway. Connected to a standard compressed air/oxygen tank or wall unit (50 psi), the device delivers a constant flow of 30 L/min and is entirely mechanical, requiring no electricity. A breath is triggered when the airway pressure (Paw) at end-expiration drops below a fixed low pressure and no longer opposes the weight of a valve that is then magnetically pulled into the inspiration position. Inspiration continues until a user-adjustable pressure above PEEP is reached, which can be set at 20–45 cm H2O on the EMX model and at 15–30 cm H2O on the HD model. Inspiration then abruptly ceases, and expiration occurs passively (secondary to lung elastic recoil) until Paw declines to 2–4 cm H2O (auto-PEEP generated by the device), which triggers the next breath. There is no capability for manual adjustment of breathing frequency (f), inspiratory time (TI), or tidal volume (VT) by the clinician. We reasoned that these parameters would be determined primarily by the patient’s respiratory system compliance (CRS) and airway resistance, a characteristic that could be exploited for monitoring purposes. In addition to this automatic mode, breaths can be delivered manually by intermittently depressing the oxygen release button on the device. The device itself has no monitoring available; however, it does provide audible feedback (ie, cycling).
A: The Oxylator EMX model (0.25 kg) and resulting flow and airway pressure (Paw) waveforms when no PEEP valve is attached. Connected to compressed gas (50 psi), the device delivers a constant flow of 30 L/min without use of electricity. Inspiration, triggered by a nadir in Paw, continues until a user-adjustable peak Paw is achieved (20–45 cm H2O). Expiration occurs passively (secondary to lung elastic recoil) until Paw declines to 2–4 cm H2O (auto-PEEP generated by default), which triggers the next breath. Breathing frequency, inspiratory time, and tidal volume are not preset and are determined primarily by the patient’s respiratory system compliance and airway resistance. The device itself has no monitoring available. B: The device with an AMBU PEEP valve attached.
Bench Study
The device was tested with a spring-loaded PEEP valve (AMBU A/S, Denmark) (Fig. 1) on a 2-chamber Michigan test lung (Michigan Instruments, Grand Rapids, Michigan) to simulate various situations of severe lung injury; the device was also tested on an ASL 5000 (IngMar Medical, Pittsburgh, Pennsylvania) to simulate unstable situations and active assisted breathing. Because of the intrinsic expiratory resistance offered by the AMBU PEEP valve, we also performed experiments with a simple and less resistive PEEP generated with an underwater column.11
The Michigan test lung was used to simulate ARDS with various respiratory mechanics during controlled ventilation. The dial pressure of the device was set at 20 cm H2O for the EMX model and at 15 cm H2O for the HD model (minimum pressure possible). CRS was varied between 15 and 70 mL/cm H2O, and airway resistance was varied between 10 and 30 cm H2O/L/s to mimic lung injury and obstructive conditions. PEEP levels were set at 0, 10, and 20 cm H2O. This resulted in > 80 combinations of CRS, resistance, and PEEP levels that were tested, and 8–10 breaths per combination were recorded.
To test the reliability of the device, we simulated the following clinical problems of abrupt onset: (1) lung volume reduction due to complete airway obstruction: starting from a CRS of 50 mL/cm H2O and a resistance of 5 cm H2O/L/s, one lung was clamped, followed by clamping of both lungs; (2) leak conditions were simulated by placing a connector with 2 small holes (2 mm in diameter) at the airway opening of the test lung, such that small leaks and large leaks were generated by opening one or both holes, respectively; and (3) pneumothorax: starting with normal CRS and resistance values, compliance of the affected lung was drastically reduced to simulate lung collapse. In addition, a heat and moisture exchanger (HME) filter was placed in-line for some measurements. To record Paw and air flow waveforms, a pneumotachograph was placed at the airway opening of the test lung and connected to a dedicated measurement setup (FluxMed GrT, MBMED, Buenos Aires, Argentina). Data were acquired at 256 Hz. For some experiments, the pressure inside the test lung (ie, reflecting alveolar pressure) was recorded. The relationships between VT, f, TI, and CRS were analyzed by linear or nonlinear curve fitting, where appropriate (see statistical analysis).
The ASL 5000 was used first to evaluate the device during assisted ventilation. With combinations of a CRS of 30 and 50 mL/cm H2O, resistance of 10 and 20 cm H2O/L/s, and a PEEP of 0 and 10 cm H2O, spontaneous breathing was simulated as an inspiratory patient effort (ie, muscle pressure) of 5, 15, and 25 cm H2O to mimic low, moderate, and high breathing effort, respectively, with spontaneous f varying between 25 and 35 breaths/min and values of mechanics simulating what has been described for COVID-19.12-14 Second, we tested the stability of the ventilation during prolonged periods (> 4 h) of unstable respiratory mechanics to simulate an unstable patient. Mechanics (similar ranges of CRS and resistance values as used for the Michigan test lung experiments) and breathing effort (muscle pressure of 3–13 cm H2O) were randomly changed every breath. Waveforms were acquired on the ASL 5000 at a sampling frequency of 512 Hz and stored for offline analyses.
Animal Model
All experimental procedures were in compliance with the guidelines of the Canadian Committee for Animal Care and were conducted after approval by the Animal Care Committee of the Research Institute in The Hospital for Sick Children. Anesthetized pigs (4 animals, 35–40 kg) were ventilated with an EMX model with dial pressure set at 20 or 25 cm H2O, using an in-line HME filter and an AMBU PEEP valve. 
was controlled with an external gas blender. The device was tested under 3 conditions for a 40-kg animal: normal lungs (CRS 26–35 mL/cm H2O), mild lung injury (CRS 15–19 mL/cm H2O), and severe injury (CRS 9–11 mL/cm H2O). Lung injury was induced by a 2-hit method: lung lavage followed by high-stretch ventilation.15 Lung lavage was performed using warmed saline 0.9%, in aliquots of 30 mL/kg until 
was < 100 mm Hg for 10 min. Pigs were then subjected to high-stretch ventilation, and CRS was measured every 15 min. High-stretch ventilation was stopped upon reaching the desired target for a drop in compliance or in 
. Mild and severe lung injury were defined according to the Berlin definition of ARDS.16
For the first 3 pigs, respiratory mechanics and gas exchange (sampled from the right carotid artery) measured during ventilation with the device were compared with conventional ventilation modes (pressure controlled and volume controlled ventilation) under conditions of normal lungs and after inducing lung injury (2 for mild injury, and 1 for severe injury). Two animals underwent conventional ventilation before changing the ventilator to the device, and one animal underwent ventilation with the device before conventional ventilation (Engström Carestation, GE, Boston, Massachusetts). The total experiment duration was 5–7 h. Because there are no standards for ventilation modes in pigs, we used pressure controlled and volume controlled modes with settings derived from previous experiments of lung injury in the laboratory to deliver lung-protective ventilation.15 PEEP and driving pressure delivered by the device (ie, peak Paw – PEEP) were matched with pressures obtained during conventional ventilation modes whenever possible (Table 1). A fourth pig was studied to assess the relationships between VT, f, TI, and CRS under conditions of normal lungs and after mild and severe lung injury. At each condition, additional changes in CRS were induced by performing stepwise lung recruitment maneuvers followed by both incremental and decremental PEEP conditions (from 0 to 25 cm H2O and back to 0 cm H2O, in steps of 5 cm H2O per 10–20 s). During all measurements, a pneumotachograph was placed at the endotracheal tube, and Paw and airway flow waveforms were acquired at 1 KHz in LabChart (ADInstruments, Sydney, Australia) and stored for offline analysis. In 2 pigs, an electrical impedance tomography device (EIT; PulmoVista 500, Dräger, Lubeck, Germany) was attached via a 16-electrode belt for additional recording of lung impedance and ventilation distribution.
Ventilatory and Gas Exchange Data From Porcine Models of Lung Injury
Statistical Analysis
Pressure and flow recordings were analyzed using software developed for Matlab R2019b (Mathworks, Natick, Massachusetts). From the flow tracings, peak inspiratory flow, f, TI, expiratory time (TE), TI to TE ratio (I-E ratio), and VT (ie, integral of inspiratory flow) were calculated. Peak Paw, total PEEP, and driving pressure were derived from the pressure waveforms. To assess the possibility of predicting VT based on respiratory timing, the relationships between VT, f, TI, set pressure, and CRS were studied using linear or nonlinear (eg, power function) regression models, where appropriate.
Stability of the device during prolonged periods of unstable respiratory mechanics was inspected visually as well as quantified as the percentage of breaths for which Paw tracings remained within expected limits. To be considered stable, the device had to meet the following criteria: (1) for a device set pressure of 20 cm H2O, driving pressure had to be at least 15 cm H2O (anticipating some resistive pressure to be subtracted); and (2) for a set PEEP of x cm H2O, the resulting total PEEP had to be between x – 1 and x + 5 cm H2O (considering additional auto-PEEP of the device).
For the animal experiments (ie, the first 3 pigs), respiratory mechanics and blood gas results for mechanical ventilation with the device and conventional modes, at different severities of induced lung injury, are presented as mean (95% CI). Modes were compared by computing the distribution (95% CI) for the difference of the means (at the .05 significance level). Data from the fourth pig were used to assess how predictable the device output (ie, the possibility to estimate VT) was; the bench-derived model fit for predicting VT using f was applied to the pig data and the percentage of breaths that fell within the 95% CI was calculated. From the EIT recordings, end-inspiratory lung impedance (EILI), end-expiratory lung impedance (EELI), and δ Z (ie, changes in impedance during the respiratory cycle) were recorded at each PEEP level; as the device’s driving pressure is constant, changes in δ Z reflect changes in compliance.17
Results
Bench Study
Basic Settings.
The device output varied as a function of the set pressure, the auto-PEEP generated by the device (2–4 cm H2O), and the respiratory mechanics. The flow was found to be constant at 28.7 ± 1.7 L/min across all conditions. A set pressure of 20 cm H2O produced a driving pressure of 15–17 cm H2O depending on the resistance and after subtracting resistive pressure and auto-PEEP of the device. Adding an HME filter reduced driving pressure by approximately 1 cm H2O, due to a small increase in resistance.
Because the flow is constant and the cycling is based on a preset pressure, the pressure limit is reached after TI depending on compliance. TI and VT vary together and depend on the set pressure and the compliance. Figure 2 shows the linear relationship between TI and VT using the EMX model with a set pressure of 20 cm H2O. The delivered VT was very close to expected based on the constant flow of 28.7 ± 1.7 L/min and the observed TI, and therefore it was highly predictable. There was some variability in the I-E ratio depending on the respiratory mechanics, ranging from 1:1 to 1:1.7. Therefore, the relationship between f (assessed over a range of 5–35 breaths/min) and VT followed a power function, and VT could be reasonably well estimated: VT = 10,530 × f−0.979, R2 = 0.95 (P < .001). During simulated severe ARDS with a set CRS of 20 mL/cm H2O, the EMX model at a set pressure of 20 cm H2O resulted in f of 30 breaths/min (range, 25–35 breaths/min) and VT of 330 mL (range, 300–360 mL) (range due to variation in simulated resistance). With a CRS > 50 mL/cm H2O, the delivered VT was larger (> 800 mL) and f was lower (< 14 breaths/min). Because VT and f change reciprocally with changes in respiratory mechanics, minute ventilation remained fairly constant across all conditions at 10.4 ± 0.3 L/min. For the HD model with a set pressure of 15 cm H2O, the driving pressure was 11–14 cm H2O. Accordingly, VT was within a clinically acceptable range (< 420 mL) for CRS < 30 mL/cm H2O. With knowledge of the set pressure, both CRS and VT can thus be reasonably well predicted from breathing frequency.
Linear relationship between inspiratory time and tidal volume delivered with the device in a bench simulation with varied compliance and resistance (R2 = 0.997, P < .001). Dashed lines represent 95% CI. Delivered tidal volume based on inspiratory time was very close to expected based on the constant flow of 28.7 ± 1.7 L/min and the inspiratory time.
Spontaneous Breathing.
Figure 3 illustrates that spontaneous ventilation was possible: inspiration was almost always triggered when breathing effort resulted in a Paw drop below total PEEP. Due to the fixed flow, the pressure curve displayed a downward curvature, suggesting relative flow starvation, which, as expected, was more pronounced with increasing inspiratory breathing effort. Wasted efforts were only seen in simulated conditions of low inspiratory breathing effort (5 cm H2O) and high compliance (50 mL/cm H2O).
Airway pressure (Paw) waveforms for simulated active assisted breathing with the device under conditions with a compliance of 30 mL/cm H2O and resistance of 20 cm H2O/L/s. A PEEP of 12 cm H2O was applied with the AMBU PEEP valve, resulting in a total PEEP of 15 cm H2O. Patient effort was simulated as a muscle pressure of (A) 5 cm H2O (low effort), (B) 15 cm H2O (moderate effort), and (C) 25 cm H2O (high effort). Note that flow starvation is present with increases in muscle pressure, as evidenced by the downward curvature of the Paw.
Simulated Acute Problems.
Simulated acute clinical problems were immediately noticed by abrupt audible changes in f (eg, inspiratory and expiratory cycling is easily audible); simulation of acute obstruction (eg, mucus plugging and atelectasis) resulted in a direct increase in f as a result of a compliance drop. In addition, the audible feedback from the device (eg, ultra-rapid cycling manifests as vibrations at high f) indicated a sudden obstructed airway. Similarly, simulation of pneumothorax resulted in increased f due to decreases in compliance. Simulation of leaks resulted in prolonged TI or audible lack of cycling with significant leaks.
Unstable Conditions.
Flow, pressure, and cycling were stable for a total of 8 h of study with different EMX models. A recording of at least 1 h was performed for each device. For a total of 11,295 simulated breaths in which CRS, resistance, and muscle pressure were randomly varied at every breath, Paw tracings (ie, total PEEP and driving pressure values) were considered within expected limits for 99.6% of breaths. The output of the studied HD model was less stable; during a 1-h unstable respiratory mechanics simulation with low CRS (15–30 mL/cm H2O), normal resistance (5–15 cm H2O/L/s), and a PEEP of 10 cm H2O (using the spring-loaded PEEP valve), 61.7% of breaths (611 of 991) had Paw values within the expected range. Visual inspection of the waveforms revealed that this was related to premature triggering (ie, triggering at pressures > 4 cm H2O above set PEEP) with rapid, small breaths.
Occasionally with the EMX model but more frequently with the HD model, premature triggering was observed when using the spring-loaded PEEP valve. It was never observed with the underwater column PEEP, nor when device was used without PEEP valve, which suggests that premature triggering was related to the way the spring-loaded PEEP valve worked and its interaction with the device. We could not identify specific respiratory mechanics settings related to the occurrence of this malfunction. Temporarily adjusting the PEEP (any change) or set pressure (by increasing it) on the device always quickly resolved the issue.
Animal Model
Device Output.
Four EMX models were studied in animals with lung injury; for any given condition, all 4 showed similar ventilation output in terms of flow and pressure waveforms. In pigs, the I-E ratio was generally close to 1:1. As a result, the expiratory flow never reached a zero-flow condition before the start of a next inspiration. Figure 4 shows the Paw output of the device and pressure controlled ventilation with CRS of 20 mL/cm H2O and PEEP of 10 cm H2O, and the comparison with device output during the bench simulation with similar respiratory mechanics and set pressure. Under matched mean Paw conditions between pressure controlled ventilation and the device, the total pressure delivered per breath (ie, area under the curve) was fairly similar despite differences in Paw profile.
Airway pressure (Paw) output of the device when tested (A) on the bench and (B) in pigs, and (C) during pressure controlled ventilation in pigs, all during conditions with a respiratory system compliance of 20 mL/cm H2O and a PEEP of 10 cm H2O. Note that in pigs, the inspiratory time to expiratory time ratio is close to 1:1. Mean Paw during pressure controlled ventilation was matched with the device, resulting in a fairly similar total pressure delivered per breath (ie, the areas under the curves are relatively similar for ventilation with the device and pressure controlled ventilation in pigs).
Gas Exchange.
Oxygenation, 
, and pH values obtained with the device and conventional modes of ventilation are shown in Table 1; values were comparable, with differences within acceptable clinical ranges among conditions and ventilator modes. For comparable PEEP and mean Paw, the EMX device achieved good VT (normal lungs: 430–513 mL; injured lungs: 250–344 mL), minute ventilation (normal lungs: 11.2–12.5 L/min; injured lungs: 10.3–11.8 L/min), 
(normal lungs: 142–236 mm Hg; injured lungs: 147–209 mm Hg), 
(normal lungs: 37–43 mm Hg; injured lungs: 52–89 mm Hg), and 
(normal lungs: 496–511; injured lungs: 252–412); values represent 95% CI of the mean; results for mild and severe lung injury were combined. At times, these values were better than those during conventional ventilation (Table 1).
Predicting VT From Respiratory Timing.
In 1 pig, a total of 9 recruitment and de-recruitment PEEP trials were analyzed during ventilation with 4 different EMX devices under 3 conditions: normal lungs, mild induced lung injury, and severe induced lung injury. Data from the resulting 1,072 breaths with a wide range of CRS (range, 5.4–46.3 mL/cm H2O) were used to evaluate the relationship between TI, f, and VT. Similar to the bench study, predicting VT using observed TI and f was possible. The equation for calculating VT from TI was VT = 487.5 × TI – 16.5, R2 = 0.99 (P < .001), and the equation for calculating VT from f was VT = 11,010 × f−0.989, R2 = 0.94 (P < .001). We then assessed the accuracy of the bench-derived formula for calculating VT based on f compared to in vivo measured values. As shown in Figure 5, 84% of the breaths fell within the 95% CI limits of the bench formula; on average, the measured VT was 2.0 ± 34.5 mL higher than the predicted VT.
Breathing frequency and tidal volume data from pigs with normal lungs and after inducing mild and severe lung injury. Additional changes in respiratory mechanics were obtained by performing PEEP recruitment maneuvers. The predictivity of the bench-derived formula for calculating tidal volume based on breathing frequency was tested by applying this formula to the pig data. As a result, 84% of the breaths (black dots) fell within the 95% CI limits of the bench formula; on average, the measured tidal volume was 2.0 ± 34.5 mL higher than the predicted tidal volume.
PEEP Titration.
A decremental PEEP trial resulted in a U-shaped response curve for f, reflecting inverse compliance changes (Fig. 6). Starting from a high PEEP, compliance increased and f decreased until a nadir of f was reached; a further reduction in PEEP resulted in increased f due to the lower compliance induced by alveolar collapse. Therefore, the lowest f corresponded to the highest CRS. Simultaneously recorded EIT data showed an inverse U-shaped pattern in δ Z (Fig. 6); because the driving pressure was constant, the highest δ Z corresponded to the highest CRS, which coincided with the lowest f. Using the hysteresis behavior method, we confirmed that this PEEP level corresponded to the highest compliance (ie, highest slope of the pressure-impedance curve); at the same PEEP level, the ventilation distribution was uniform between ventral and dorsal regions (Fig. 6).
After lung injury, decremental PEEP steps were conducted while using the Oxylator EMX device. (A) Paw and breathing frequency. As the applied PEEP was decreased from 25 cm H2O to 0 cm H2O, the measured PEEP decreased from 30 cm H2O to 3 cm H2O. Breathing frequency follows a U-shaped response curve with decreasing PEEP. (B) Data from a different pig while performing a decremental PEEP trial. In addition to the U-shaped response curve seen in breathing frequency, a reverse pattern is seen with δ Z measured with EIT. Because the driving pressure is constant, highest δ Z corresponds to best lung compliance, which in turn corresponds to lowest breathing frequency. (C) Lung impedance numbers at inspiration and expiration during the decremental PEEP steps along with corresponding EIT images in the same pig as in (B). The best compliance is in the middle steeper portion of the curve and corresponds to the best PEEP in (B), which is at 14 cm H2O. (D) Ventilation distribution for the dorsal and ventral lung regions, showing a uniform distribution at this best PEEP level. At lower PEEP levels, there is a discrepancy in ventilation distribution between ventral and dorsal lung regions, due to alveolar collapse occurring in the dorsal lungs. The uniform ventilation distribution at higher PEEP levels (> 14 cm H2O) is the result of hyperinflation as shown in (B) and (C). Paw = airway pressure; EIT = electrical impedance tomography; EELI = end-expiratory lung impedance; EILI = end-inspiratory lung impedance; δ Z = δ impedance.
Discussion
In the era of the COVID-19 pandemic and in preparation for the next pandemic, we need simple, efficient, and safe ventilation devices to meet the unprecedented ventilator demands of patients, and to cope with extraordinary logistical challenges. Our evaluation of the Oxylator yielded 4 notable results. First, for severe lung injury (low compliance), the device offers simplicity, electricity independence, and portability, and it delivers a VT within a clinically acceptable range at a relatively high f. In normal lungs, VT will be larger.18 This makes the device especially suitable for respiratory systems with low compliance and short time constants. Second, the inability to adjust ventilator parameters other than the dial pressure makes controlling ventilation challenging; however, due to the predictable output and with monitoring of respiratory timing (via an external monitoring device or audible feedback), the user can calculate CRS and VT very precisely from TI and with reasonable accuracy from the f (given variations in the I-E ratio). Monitoring of respiratory timing is advisable because it could also help in recognizing acute clinical problems at the bedside. Third, the response curve in f during a decremental PEEP trial confirms the dependence of f on CRS and could help the clinician to adjust the PEEP level. Finally, gas exchange in an animal model of lung injury was similar to conventional ventilation from a standard mechanical ventilator, using currently recommended pressure or volume-prioritized ventilation settings.
Device Performance
The ongoing pandemic has put a burden on critical care resources, especially mechanical ventilators. While modern ICU ventilators are sophisticated and reliable and provide extensive patient monitoring, they are expensive, making stockpiling of these ventilators cost-prohibitive. ICU ventilators are complex devices requiring extensive training for safe use and maintenance, and they require sources of electrical power, air, and oxygen pressure. They are not portable, and they require extensive quality control (especially if produced in non-standard conditions) and a reliable supply of disposable equipment.7,19 Ultra-simple and user-friendly, durable, portable, and low-cost ventilation solutions are therefore potentially important.1-3,7 The Oxylator could meet these requirements. It is currently used worldwide in prehospital settings during rescue situations such as after cardiac arrest or severe trauma.9,10,18 However, the device could be recommended for use in acutely ill patients with lung injury if safe, effective, and individualized ventilation can be achieved to deliver protective ventilation. Over a wide range of respiratory mechanics conditions, we demonstrated stable device output, successful delivery of PEEP, and protective VT in mild to severe lung injury, but, paradoxically, less in conditions of normal compliance, at least using the lowest pressure of the EMX model. Furthermore, short-term use of the EMX model resulted in ventilation and gas exchange comparable to conventional ventilation modes. This suggests that the device could be a relatively safe and efficient solution for short-term rescue ventilation.
Breathing Frequency as Minimal Monitoring
In contrast to conventional ventilation modes, the pressure cycling mechanism makes the VT delivered by the Oxylator directly dependent on the patient’s respiratory mechanics. Because the device lacks monitoring displays, challenges arise and monitoring would be necessary to avoid excessive or insufficient VT. Our results indicate that bedside estimation of VT is possible based on respiratory timing. Changes in f could predict VT using a bench-derived formula that was validated in an animal model over a large range of respiratory mechanics conditions, showing clinically acceptable variability in predicted VT. This formula could be incorporated into an online calculator. In addition, monitoring f allows recognition of changes in respiratory mechanics due to abrupt clinical problems (eg, acute obstruction, leaks) or disease progression. Furthermore, we suggest a method for setting PEEP at the bedside based on the f response to changes in compliance during a decremental PEEP trial. Indeed, the PEEP level based on this response corresponded to the PEEP related to the best EIT-derived compliance, as confirmed with lung impedance EIT data at this pressure level and uniform ventilation distribution. Although f could be monitored via the audible feedback of the device, this requires a caregiver to be present to listen carefully to the device’s cycling. External devices such as a pulse oximeter, end-tidal CO2 monitor, or portable gas flow analyzer placed in series with the device would therefore be recommended.
Limitations and Future Work
Some limitations regarding the use of the device should be addressed. First, without an external blender, the gas is pure oxygen, which limits the duration of use. However, an air/oxygen blender could be added to control 
delivery, although this may affect portability of the device and increase cost. Second, the minimum driving pressure of the EMX model limits its use in lungs with normal compliance. In these patients, excessive VT can be prevented when using the device in manual mode; however, this method requires a clinician to operate the device. Although lower driving pressures can be achieved with the HD model, fewer devices of this model are available worldwide compared to the EMX model (> 20,000 devices have been sold). It should also be noted that, according to its intended use as an automatic resuscitator, the device should not be used unattended. Third, the use of a spring-loaded PEEP valve occasionally resulted in device malfunction, which manifested as premature triggering with high f. This problem never occurred with non-resistive underwater PEEP generation. When this problem occurred, it was always easily resolved quickly by temporarily changing the PEEP level or the set pressure of the device; recognizing this problem requires monitoring of f. Last, spontaneous breathing is possible when using the device with added PEEP; however, due to the fixed flow, this may become uncomfortable if high respiratory efforts are allowed. Use without a PEEP valve (eg, during noninvasive application when connected to a face mask) allows ambient air to be inhaled when patient effort is high, but this should be used with caution and when PEEP is not necessary. Addressing these technical limitations in a next-generation model would enhance future clinical implementation. In addition, before widespread use of the device, its reliability and safe patient monitoring should be verified in a clinical setting of severe lung injury.
Conclusions
In bench and porcine models of severe lung injury, the Oxylator is a simple automated resuscitator delivering stable ventilation with VT within a clinically acceptable range in low compliance lungs. External monitoring of respiratory timing is advisable and allows estimation of the delivered VT, could guide setting PEEP, and helps with recognizing sudden changes in respiratory mechanics due to clinical problems or disease progression. In resource-constrained settings such as those seen during the coronavirus pandemic or in environments with limited equipment, the Oxylator can be a low-cost practical rescue solution for providing short-term ventilatory support as a temporary bridge until an ICU ventilator becomes available.
Footnotes
- Correspondence: Laurent J Brochard MD, Li Ka Shing Knowledge Institute, 209 Victoria St, Room 4-08, Toronto, ON, M5B 1T8 Canada. E-mail: laurent.brochard{at}unityhealth.to
See the Related Editorial on Page 533
Drs Dorian and Brochard are co-senior authors.
This work was supported in part by Toronto COVID-19 Action Fund from the University of Toronto and CIHR (FDN143285 and OV3-170344). The Oxylator devices were provided free of charge by CPR Medical Devices; the company played no role in the design and conduct of the study.
- Copyright © 2021 by Daedalus Enterprises
