Analysis of the fraction of delivered oxygen by noninvasive ventilation devices working as invasive ventilators in pandemic times


 Purpose: In the context of the COVID-19 pandemic, mandatory ventilation with noninvasive ventilation (NIV) devices is a valid option when intensive care/anesthesia ventilators are unavailable. The fraction of delivered oxygen (FDO2) by NIV devices in intubated patients is unknown.Method: We simulated intubated patients with normal and sick lungs. NIV was used in pressure control mode with protective lung ventilatory settings. O2 flow was added into the NIV circuit in incremental steps of 1 L/min (from 1 to 15 L). The FDO2 in breathing gases was measured by a paramagnetic O2 sensor placed behind the endotracheal tube. Three NIV circuit options were analyzed: 1) leak at HME filter close to the patient, 2) anesthesia Bain circuit with leak distal to the patient, and 3) leak throughout a non-rebreathing valve near the patient.Results: FDO2 increased proportional to the supplemental O2 flow in all NIV options and in both kinds of patients. The range of FDO2 came from 0.25 to 0.98 in both, healthy and sick lungs. At 5 L/min, FDO2 was 0.53±0.04 and 0.47±0.02 in option 1 and 0.53±0.04 and 0.47±0.02 in option 2 for healthy and sick lungs, respectively. In option 3, 5 L/min of O2 reached 0.84±0.08 in healthy and 0.74±0.09 in sick lungs. Conclusions: In all setups, FDO2 was proportional to the administered O2 flow and it covered the range of FDO2 values commonly observed in ventilated patients.


Introduction
Respiratory failure induced by the SARS-CoV-2 virus is a key feature in most patients with COVID-19 [1,2].
The high contagiousness of the virus (Ro = 2.2-3.5) [3] increases the number of people infected exponentially, explaining why health systems collapsed all over the world. The problem is particularly severe in intensive care units (ICU) where beds and ventilators could be fully occupied with  cases that require invasive ventilation longer than usual. In such critical scenario even anesthesia workstations have been utilized to ventilate COVID-19 patients, limiting the number of this crucial equipment in the operating rooms.
The use of noninvasive ventilation (NIV) devices in-lieu of invasive ICU ventilators has been recently suggested in the context of the COVID-19 pandemic [4]. Noninvasive ventilation was proposed as a second line strategy when ventilators and/or anesthesia machines are not available and patients need ventilatory support for acute respiratory failure or for general anesthesia. This is clearly an emergency situation for the pandemic that is not appropriated for normal circumstances. The FDA approved NIV devices to be used in intubated patients at the time of writing of this manuscript (https://www.fda.gov/medical-devices/letters-health-care-providers/ventilator-supply mitigationstrategies-letter-health-care-providers. Accessed April 20, 2020).
There are many comparative advantages and disadvantages between NIV devices when compared to ICU and anesthesia workstations [5]. Perhaps the main limitation of NIV is the di culty to get precise fraction of delivered oxygen (FDO 2 ), therefore, speci c O 2 ow/ FDO 2 tables are needed to adjust the oxygen therapy. Such information was described in two studies that simulated NIV in spontaneous breathing patients using masks [6,7]. However, these data are only approximations because the FDO 2 is easily modify by unintended leaks around poor-tting masks and the patient's minute ventilation. In intubated patients with cuffed endotracheal tubes, NIV devices work with an intended air-leak in the ventilatory circuit but without any additional "unintended" leaks like with facial interfaces. Thus, for most of the NIV devices, the FDO 2 delivered to patients undergoing mandatory ventilation is unknown.
The main objective of this brief report was to analyze the FDO 2 supplied by NIV devices working as invasive ventilators. The analysis was performed in a bench study at this rst step, simulating patients with healthy and sick lungs using different NIV circuit con gurations at different supplemental O 2 ows.
End-point of the analysis was the measurement of FDO 2 at the airways opening. The primary outcome was to describe the FDO 2 obtained at different additional O 2 ow for different NIV circuit dispositions.
The IRB of the Hospital Privado de Comunidad, Mar del Plata, Argentina approved the use of clinical data recorded in three anesthetized patients to illustrate the results of simulations (waiver for patients' written informed consent).

NIV devices working as an invasive ventilator
To operate as an invasive ICU ventilator the NIV device must have the following features: To allow pressure control and/or S/T modes.
The NIV device must reach an inspiratory positive airway pressure (IPAP) ≥ 30 cmH 2 O and an expiratory positive airway pressure (EPAP) ≥ 10 cmH 2 An obligated gas leak port to washout CO 2 from the single limb NIV circuit.
An external O 2 source supply. We have administered O 2 through a port placed between the singlelimb circuit and the NIV device.
A heat and moisture exchange (HME) antibacterial/antiviral lter must be placed between the endotracheal tube (ETT) and the NIV circuit. The HME prevents SARS-CoV-2 virus dispersion and keeps humidity of inhaled gas.
Three NIV circuit con gurations were proposed to implement ventilation in intubated patients ( Figure 1) [4]. These are circuit con gurations described for such altered standard of care that must be only used in the context of COVID-19 pandemic. The rst option is the standard circuit con guration that has a leak at the HME lter, nearest to the patient. The second one is a modi ed Bain system ideated to perform general anesthesia, which has the advantage to keep heat and humidity in the breathing gases [8]. The circuit has double coaxial tubes one inside the other. The outer tube is connected to the NIV device and delivers inspired gases to the patient while the inner tube transports expired gases to the ambient (leak closer to the NIV equipment). The third option has a non-rebreathing valve like Ruben or Duckbill (similar to those found in Ambu® resuscitation bags) [9], which delivers gases to the patient and then eliminates gases to the ambient throughout a PEEP valve. This atypical circuit option has the theoretical advantage to deliver more O 2 and to decrease CO 2 re-breathing.

Simulations
The analysis was done in the Simulation Center of the Buenos Aires Association of Anesthesia, Analgesia and Reanimation. Data was collected by the ASL 5000 Breathing Simulator (IngMar Medical, Pittsburgh, USA), which was connected to the NIV device (Stellar 150, ResMed Inc., Sydney, Australia) by a cuffed nº 8 endotracheal tube. This device can be used for noninvasive and invasive ventilation according to manufacturer's speci cations. Respiratory mechanics and CO 2 -O 2 signals were obtained with sensors placed at the airways opening (S5 device, GE Healthcare/Datex-Ohmeda, Helsinki, Finland). The O 2 was measured by paramagnetic sensor with an accuracy < 2% of reading, rise time < 260 milliseconds and measurement range between 0 to 100% [9,10]. Sensors were calibrated before protocol as described by the manufacturer. Data was recorded by the software Datex Collect (GE Healthcare/Datex-Ohmeda, Helsinki, Finland) in a laptop and was analyzed off-line.
We simulated two kinds of patients based on lung mechanics. One patient with healthy lungs, with a respiratory compliance of 50 mL/cmH 2       The values of FDO 2 obtained by simulations (Table 1) corresponded to the data observed in anesthetized patients at similar ventilatory settings and supplemental O 2 ow (Table 2). However, the FDO 2 values we found are much higher than those simulations reported during spontaneous breathing using facial interfaces [6,7]. These differences are probably due to the unintended leaks commonly observed around masks that do not happen in patients with sealed circuit and cuffed endotracheal tube. The leaks throughout the mask are compensated by the NIV device increasing ow to maintain the set pressures.
This effect dilutes the O 2 and changes FDO 2 within the NIV circuit because the supplemental O 2 ow keeps constant.
The lower FDO 2 observed in sick lungs compared with healthy lungs could be explained by the high IPAP and EPAP used in the former. These results t with the description of Schwartz et al. [6] and Thys et al. [7] using bench models and volunteers. The authors found that selecting high pressures increase not only the ow through the leak but also the NIV ow to compensate such leaks. The nal result is the dilution of O 2 within the NIV circuit and the consequent reduction in the FDO 2 .
The NIV circuit with the unidirectional, non-rebreathing valve showed the highest FDO 2 . This is because such valve allows unidirectional ow during the respiratory cycle phases. The "leak" is closed during inspiration and fresh gases reach the patient. Then, the expired gases rich in CO 2 tend to scape outside through the valve during expiration. The combination of these effects is a lower dilution of O 2 within the single circuit of NIV. Thus, this NIV circuit disposition must be used in those cases when high FDO 2 is necessary like in patients with severe respiratory failure or during one-lung anesthesia. The main disadvantage of this NIV circuit disposition is the additional expiratory work imposed by the nonrebreathing and PEEP valves. This is a problem during assisted ventilation (S/T mode) that can be easily resolved by changing to circuit options 1 or 2 at the very moment when a patient is ready to re-start spontaneous breathing.

Limitations
We have simulated changes in FDO 2 caused by increments in the O 2 ow supply, keeping constant all the other ventilatory variables. Such variables like IPAP/EPAP values, placement of circuit leak, place of supplemental O 2 administration, respiratory rate, inspiratory time, etc. could affect the FDO 2 in NIV devices used as invasive ventilators. Due to the urgency to get alternative ventilatory options during the COVID-19 pandemic, we believe that the information we generated, while limited to changes in additional O 2 ow, it may prove useful to intensive care health-providers and anesthesiologists at this stage. Further studies should be done to get more information on the topic.

Conclusion
This study gives clinical information to manage FDO 2 when NIV devices are used for mandatory ventilation in intubated patients. The range of FDO 2 reached by all NIV circuit options covers the range of values used in our ventilated patients with intensive care/anesthesia ventilators. Our results should not be transferred to patients in spontaneous assisted ventilation and surely new studies should be done to cover this issue.

Declarations
Informed Consent: The IRB of the Hospital Privado de Comunidad, Mar del Plata, Argentina approved the use of clinical data recorded in three anesthetized patients to illustrate the results of simulations (waiver for patients' written informed consent).  Three NIV circuit con gurations to perform mandatory ventilation were studied. In option 1 the leak is placed distal to the NIV device, at the heat and moisture exchange (HME) lter. The second option is a modi ed anesthesia Bain coaxial circuit, where inspired gases go through the external limb and the expired gases are eliminated by the internal coaxial tube. The third option has a non-rebreathing unidirectional valve. Such valve is placed behind the HME where all the exhaust gases come outside the circuit through a PEEP valve. In all options, supplemental oxygen (O2) is administered between the NIV device and its single circuit. Flow, pressure, capnography (CO2) and oxygraphy (O2) temporal series were collected in a patient undergoing general anesthesia using a Bain circuit (Option 2). Note that oxygraphy is a mirror image of capnography.
Page 13/14 Supplemental O2 ow was changed in one anesthetized patient during xed ventilation using option 1 (leak at HME lter). Ventilatory settings was similar to the one simulated in a healthy patient. The gure on the top depicts how fast the fraction of O2 at end-inspiration and end-expiration reached stable values after the addition of 5 L/min of O2.