Non-invasive ventilation (NIV) is the standard of care for the treatment of chronic hypercapnic respiratory failure.1 It can be conducted through different configurations, single-limb circuits with facial masks being the most frequently met in the home setting.1 This implies the presence of an intentional leak either at the mask or between the mask and circuit, preventing exhaled air to be rebreathed. However, substantial dissemination of exhaled particles into patients’ environment has been documented.2

With the current COVID-19 pandemic, growing interest regarding the risk for NIV to promote airborne transmission has emerged.3 This concern has led by itself to discuss the indication of this therapy in patients requiring hospital admission for de novo SARS-CoV-2 infection or in situations of acute-on-chronic respiratory failure.4–6 In the latter situation, patients treated with chronic NIV may worsen with ventilation cessation and protective measures allowing for pursuing therapy in and out of the hospital setting are necessary.

Thus, several adaptations of circuit configuration in order to reduce the risk of NIV-related viral contaminations have been proposed worldwide.7–11 Most of these recommendations share the common aspect that they increase instrumental dead space (VD) and airway resistance compared to “usual” configuration. As a consequence, any modification of NIV circuit configuration when applying these protective recommendations may require NIV settings adjustments to preserve ventilation efficiency.

We therefore designed this bench study in order to evaluate the effects of circuit configurations recommended for NIV during COVID-19 pandemic on the pressure support (PS) and expired tidal volume (VTE) modifications required to maintain treatment efficiency with respect to end-tidal CO2 partial pressure (PETCO2).

For configuration A, reference PETCO2 at steady state was 38(0)mmHg, with a PS set at 10cmH2O resulting in a VTE of 432(2)mL. With the baseline ventilatory parameters, substantial variations in VTE were observed, associated with a clinically relevant increase in PETCO2 (Table 1; iso-PS variables). Consequently, all the configurations evaluated required higher PS to maintain stable PETCO2 (Table 1; iso-PETCO2 variables). V′CO2 remained within clinically acceptable range (180(5)mL/min) throughout the protocol. The absolute increase in PS (ΔPS, cmH2O) to maintain iso-PETCO2 for each configuration is depicted in Fig. 1, lower panel. This resulted in higher VTE for all configurations evaluated compared to reference configuration (Fig. 1, lower panel), respectively +142(7), +91(1), +216(1), +79(2) and +144(7)mL with configurations B–F. The calculation of additional VD according to Bohr's equation yielded similar results, with an increase in VD by +143(7), +91(2), +216(2), +81(3) and +148(6)mL with configurations B–F, respectively. Ventilatory parameters measured in the condition of stable alveolar ventilation (iso-PETCO2) are displayed in Table 2. The time course of Paw, V′aw and PCO2 for the reference configuration (A) and the configuration requiring the maximum PS and VTE modifications to reach reference PETCO2 (configuration D) is depicted in Fig. 2.

Fig. 2.

Time course of Paw, V′aw and PCO2 for configurations A and D.

Paw, airway pressure; V′aw, respiratory flow; PCO2, partial pressure of CO2. Vertical purple lines: solid line, beginning of inspiration; dashed line, end of inspiration. The figure illustrates the need to increase pressure support and thus tidal volume to maintain isocapnia (38(1)mmHg). Note that the duration of PCO2 decline during inspiration in configuration D is much longer than configuration A, suggesting higher volume of CO2 rebreathed from additional instrumental dead space.

(0.2MB).

Compared to reference (config. A), circuit compliance was increased with all the configurations evaluated (see supplementary Table 1). Mean compliance ranged from 0.4 to 2.2mL/cmH2O with configurations A–F, respectively. At 0.5L/s, the lowest (1.4cmH2O/L/s) and highest (3.3cmH2O/L/s) inspiratory resistances were encountered with configurations C and F, respectively. Configurations B, C and D resulted in substantially higher expiratory resistances than configurations A, E and F (see supplementary Table 1), with the lowest (2.2cmH2O/L/s) and highest (20.0cmH2O/L/s) values encountered with configurations E and D, respectively. Supplementary Fig. 2 displays the pressure–flow relationship for each configuration during inspiration and expiration.

In this study, we demonstrated that substantial adjustments in NIV settings are needed to maintain alveolar ventilation when modifications of circuit configuration are required to prevent NIV-related contaminations in the current context of COVID-19 pandemic. This is particularly relevant in patients chronically ventilated pursuing their therapy at home, or exhibiting acute-on-chronic exacerbation requiring hospital admission.7,10,14 Of note, even the change from configuration A–E which is often made when patient's condition requires switching to a life-support ventilator needed setting adjustments.

In contrast, this finding is less relevant in patients for whom NIV is initiated during an acute episode because treatment efficiency can be expected, in this context, to be carefully monitored and settings adapted whenever necessary. Moreover, in patients hospitalised for an acute lung injury related to coronavirus, the initiation of bilevel NIV is highly controversial.15–17 Most guidelines recommend NIV in patients with SARS-CoV-2 infection as a “default” option when oxygen therapy or constant positive airway pressure (CPAP) are not sufficient to improve patient's status, provided that appropriate personal protective equipment is available.10,14,18,19 Its indications in this context therefore remain restricted to situations in which access to invasive mechanical ventilation is limited, in patients not eligible to ICU admission, or in patients with underlying respiratory disease or previously undergoing long-term NIV in which NIV cessation could worsen even more their condition and for whom the results of the present study apply as well.14,20

Unexpectedly, we found discrepancies between values set on the ventilator and values measured between the mannequin head and the test lung even for baseline configuration (A) in which PEEPaw was 1.4cmH2O above EPAP setting. Of note, as most ventilators do not control the pressure directly inside the mask, any additional material increasing resistances between the point of pressure measurement and the mask may affect the delivered pressure inside the mask during NIV. Advices from ventilators’ manufacturers could be helpful in daily practice when unusual circuit configurations are considered in patients undergoing NIV.

In our experimental setup, the increased PEEPaw could at least in part be explained by the elevated RR (25bpm), which however is in line with clinical characteristics of severe patients and may promote dynamic hyperinflation (Fig. 2).21 This phenomenon was even more pronounced with other configurations (>3.5cmH2O with configurations D and F). From a theoretical point of view, increased expiratory resistances related to additional equipment (filters, tubing) might explain this observation. However, the highest PEEPaw monitored for all configurations was encountered with configuration F which actually was one of the configurations providing the lowest expiratory resistance (Supplementary Fig. 2). Conversely, configuration D yielded similar PEEPaw as configuration F, despite substantially higher expiratory resistances. It should be noted that for single with leak circuits (configurations A, B, C and D), expiratory resistances were measured with the circuit closed at ventilator's side, so that exhalation could only occur through the intentional leak. This may not accurately reflect clinical situations in which an important proportion of expired tidal volume may occur inside the circuit despite the presence of an intentional leak and contribute both to increasing CO2 rebreathing and reducing expiratory resistance as previously demonstrated.22 Either way, a simple regression analysis showed a significant linear relationship between tidal volume and PEEPaw (R2=0.76 and P=.014) whereas no relationship was detected between expiratory resistance and PEEPaw. This suggests that increased PEEPaw observed in the condition of iso-PETCO2 was most likely driven by the increase in VT (additional instrumental dead space) contributing, along with high respiratory rate, to hyperinflation of the model, rather than variations in circuit resistances.

Our study has several limitations.

First, in our experimental setup, driving pressure was generated by the device solely. Simulating patient's effort would have interestingly given the opportunity to evaluate the effects of circuit configurations with CPAP mode and to assess the variations in respiratory effort required to preserve alveolar ventilation in these different situations. Unfortunately, using one chamber of the Michigan test lung as a driving chamber to simulate patient's effort would have substantially reduced the end-expiratory lung volume of the model. With clinically relevant V′CO2, a parallel connection of the two chambers of the test lung was required to reach relevant PETCO2. Another limitation to the evaluation of CPAP mode is that this study aimed to test all the circuits modifications proposed in the literature to prevent coronavirus contamination in subjects chronically ventilated, both with single and double limb circuits. Almost all adult patients treated with CPAP mode in the context of domiciliary usage are treated with devices with single with leak configuration only. The external validity of evaluating double limb circuit modifications (configurations E and F) in subjects undergoing long term CPAP would have been broadly questionable. Additionally, even though the use of CPAP mode has been widely described in COVID-19 patients,19,21 this ventilatory mode with constant pressure would not have allowed to determine the variations of driving pressure required to maintain isocapnia with respect to the design of our bench and would also not have reflected the population of interest of our study. The results of the present study therefore do not apply to this mode. Finally, we therefore decided to use a controlled ventilation condition without patient effort in order to simplify the model, considering that modifications observed in this context were only attributable to the modifications of instrumental dead space and resistances that we wanted to evaluate. Consequently, we adjusted PS according to PETCO2. In clinical practice it can be expected that patients, as long as their muscle capacities allows to do so, would increase their work of breathing – and obviously V′CO2 – in order to regulate PCO2, as previously demonstrated with heat and moisture exchangers.23 Instead of continuous monitoring of PETCO2, adjusting PS according both to clinical signs of respiratory effort and to tidal volume would be less cumbersome and more appropriate in clinical routine.24 Nevertheless, the aim of our bench study was limited to a simple comparison in tidal volumes keeping PETCO2, respiratory rate and V′CO2 constant in order to consider that, in these conditions, the tidal volume differences would therefore represent the instrumental dead space differences according to Bohr's equation.25

Second, we used a viral/bacterial filter with a large internal volume (120mL). Even though this reference was found to be used by several other teams, such large internal volume deserves discussion.8,26 Along with the use of a non-vented mask, this resulted in an increase in VTE required to reach isocapnia, i.e. in an increase of instrumental dead space, of 142(7) and 144(7)mL with configurations B and F, respectively, and in a PS (and in our situation driving pressure) nearly twice as big as with configuration A. Of note, because of fluid dynamics variations depending on circuit configuration (and particularly on the position of intentional leak), the internal volume of instrumental dead space in mL may not accurately reflect the absolute value of additional dynamic dead space as previously demonstrated by Saatci et al.27 Nevertheless, as far as equipment is available, any reduction in instrumental dead space should be encouraged in order to reduce respiratory effort and prevent patient self-inflicted lung injury (P-SILI), especially in the context of acute lung injury related to COVID-19.28

Large differences in additional VD exist between the recommended configurations. It is noteworthy that the configuration which induces the lowest increase of VD with single-limb circuit is configuration C, which proposes to laterally position the leak and filter with a T-piece placed immediately after the mask.8 This allows a bypass where the expiratory gas is blown in this lateral space, preventing from its reinhalation during the next insufflation. The need to re-evaluate treatment efficiency and settings is crucial whenever protective measures influencing NIV equipment are considered.

All authors contributed substantially to study conception and design; MD, KL, BL and FL contributed to data acquisition; MD, GB, BL and FL contributed to data analysis and interpretation; MD, GB, ML, BL and FL drafted the manuscript. All authors revised the manuscript for important intellectual content, approved the final version, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

None declared.

MD is a former employee of ResMed SAS, which had no involvement in the current study. Other authors have no conflicts of interest to disclose.