Noninvasive ventilation (NIV) is extensively used among patients with acute and chronic respiratory failure, reducing the intubation rate, the risk of respiratory failure after extubation, hospital length of stay, and mortality (1–5). In certain patient populations, such as those experiencing an exacerbation of chronic obstructive pulmonary disease (COPD), NIV is considered first-line therapy and the standard of care (1–3).

NIV is delivered via a noninvasive ventilator with a single limb, utilizing an exhalation port that directs exhaled gasses from the breathing circuit to the atmosphere. Currently, four types of exhalation valve are available for clinical practice: the single-arch, plateau, whisper swivel, and vented mask (integrated leak port within the mask) valves (6). The characteristics of such valves including design features that allow controlled amounts of air to escape as “intentional leaks” vary greatly (7), influencing rebreathing risk and carbon dioxide retention, particularly in hypercapnic respiratory failure. Recent clinical data (8) suggested that insufficient CO₂ clearance may be associated with worse medium-term outcomes, highlighting the importance of CO₂ washout during inpatient NIV use. Besides, the characteristics of exhalation valves affecting the ability of a ventilator to generate pressure and thus its performance; this is particularly true of plateau exhalation valves (7). However, it remains unclear whether commercial exhalation valves differ in their effects on CO₂ rebreathing and ventilator performance when used with different ventilators, and the optimal standards for exhalation valve selection have yet to be identified. We therefore employed bench methodologies to systematically quantify valve performance mechanisms, establishing a mechanistic basis for subsequent clinical validation.

In the present bench study, we applied the simulated lung model to comprehensively compare CO₂ rebreathing among the four types of valve in combination with three noninvasive ventilators. We also evaluated the impact of valve type on ventilator performance, including triggering, control, and monitoring parameters.

We comprehensively assessed the effects of four types of exhalation valve on CO₂ rebreathing and ventilator performance under different NIV settings and ventilators. Although no significant differences in CO₂ rebreathing were observed among the single-arch, whisper swivel, and plateau valves, all showed significantly higher rebreathing volumes than the mask valve. Additionally, except for the V60 ventilator, some ventilator performance parameters were significantly lower for the plateau valve than for the others, and the rate of difference in V_T between the ventilator and the simulated lung with all exhalation valves exceeded 10%.

NIV can be delivered via a single-limb circuit, requiring a valve to enable exhalation. Clinically, four types of such valve are commonly used, depending on preference and equipment availability. However, their structural and functional characteristics differ. Previous study has indicated that leakage from the single-arch, whisper swivel, and mask valves increases with pressure, whereas the plateau valve maintains relatively constant leakage across different pressure levels due to its unique silicon diaphragm structure (7). Our study corroborates these findings, and further comprehensively analyzed their impact on CO₂ rebreathing and ventilator performance.

Both bench and clinical studies (14–18) have demonstrated that CO₂ may not be adequately cleared during NIV, for an additional rebreathing process occurred and exhaled gasses constituted a fraction of the delivered tidal volume. Various factors—including patient-associated characteristics (e.g., baseline CO₂ levels, spontaneous breathing patterns), ventilator settings (e.g., IPAP, EPAP), and interface-related elements (e.g., exhalation valve type and position)—have been reported to affect CO₂ rebreathing during NIV (15–18). The simulated COPD lung model, which replicates key pathophysiological features of hypercapnic COPD such as increased dead space, has been widely used in previous bench studies (9, 11) to evaluate CO₂ clearance efficacy. Similarly, maintaining a simulated PetCO₂ of 80 mmHg—consistent with our previous approach (9)—provides a standardized hypercapnic condition for assessing the impact of different exhalation valves on rebreathing. As in prior methodology (15), the calculated CO₂ rebreathing represented the total rebreathing volume from the ventilator circuit, mask, and upper airway dead space of the mannequin, and not solely the CO₂ rebreathed from the circuit. Besides, previous studies also suggested that CO₂ rebreathing can be minimized by using an EPAP of up to 8 cmH₂O or an IPAP of up to 20 cmH₂O (16). Therefore, to eliminate ventilator parameter influences while accounting for clinical constraints (specifically, the 4 cmH₂O minimum EPAP setting on most devices), we applied IPAP levels of 12 or 16 cmH₂O with a fixed EPAP of 4 cmH₂O in this bench study. This design enables isolated investigation of exhalation valve effects on CO₂ rebreathing within an in vitro COPD model.

We found that CO₂ rebreathing with the mask valve was significantly lower than with the single-arch, whisper swivel, and plateau valves. This aligns with Schettino et al. (15), who demonstrated that exhalation port position influences CO₂ rebreathing. Compared to circuit-located leak ports, the mask-integrated exhalation port (as in the Facial-MEP design) enhances residual gas washout by fresh airflow, reducing dead space ventilation and rebreathing volume. Meanwhile, our previous work confirms substantial design-dependent variation in CO₂ rebreathing among oronasal masks with integrated ports, with mean values ranging 11.01–16.78 mL (9). Notably, all integrated-port masks exhibited lower rebreathing volumes than single-arch, whisper swivel, and plateau valves in this study.

Among the circuit-located exhalation ports, the plateau valve is generally believed to facilitate lower CO₂ rebreathing due to its unique controlled leakage mechanism (16). However, no significant differences in CO₂ rebreathing were observed among the single-arch, plateau, and whisper swivel valves in our study. This aligns with Hill et al.’s in vivo findings (19), where plateau valves showed no clinical advantage over whisper swivel valves in long-term NIV patients. This null effect likely relates to leakage characteristics. Except the minimum EPAP (4 cmH₂O), plateau valves demonstrated equivalent leakage to whisper swivel and single-arch valves at both 12 and 20 cmH₂O (Figure 2). Consequently, at our tested IPAP levels (12/16 cmH₂O) with fixed EPAP (4 cmH₂O), plateau valves conferred no additional rebreathing reduction. Notably, Ferguson and Gilmartin (16) observed superior performance of plateau valves at EPAP = 0 cmH₂O, with valve differences disappearing at higher EPAP levels. Thus, at clinically relevant pressures (>4 cmH₂O EPAP), rebreathing differences among these circuit-located valves become negligible. Further integrated bench-to-bedside studies are warranted to define valve selection criteria across therapeutic pressure ranges.

In terms of ventilator performance, we found that the plateau valve differed from the others. Triggering parameters (Ttrig, Tpmin, Ptrig) and control parameters (T90%, Ti) were significantly lower than for the others across the tested ventilators, except for the V60 ventilator. This may be due to the constant leakage of the plateau valve, and there is a greater leakage during the expiratory phase when EPAP is relatively low. This greater leakage may cause the ventilator to mistakenly perceive that the patient starting to inhale, thereby prematurely triggering the ventilator to deliver gas to the patient, resulting in a decrease in triggering parameters. Additionally, the relatively low leakage with the plateau valve at higher IPAP settings (16 cmH₂O) may cause the ventilator to switch to expiration earlier, resulting in shorter inspiratory times and reduced tidal volumes. Therefore, when selecting exhalation valves in clinical practice, it is crucial to consider their impact on triggering and control performance during NIV.

Furthermore, we found that the rate of difference in V_T between the ventilator and the simulated lung exceeded 10% for all exhalation valve types when using the Flexo and Stellar150 ventilators but not the V60 ventilator. During NIV, the monitored tidal volume displayed on the ventilator is an estimated value. Luján et al. (20) reported that different valve types have different leakage, which can affect the accuracy of ventilator parameter monitoring during NIV. To be able to display accurate tidal and minute volumes, the ventilator must know the intentional leak characteristics of the specific mask/patient interface and exhalation port. With V60 ventilator, select the desired exhalation port type or/and run an exhalation port test is needed before the ventilator starts working. We previously found that tidal volume monitoring was more accurate after leak tests when using certain industrial masks (13). These findings suggest that exhalation valves may not be interchangeable among ventilators with single-limb circuits. If interchangeable, leak tests of the valves from different manufacturers may be required.

Our in vitro findings yield some clinical and research implications. Consistent with prior studies, mask-integrated valve demonstrated significantly lower CO₂ rebreathing than circuit-located valves (single-arch, whisper swivel, plateau) during NIV, aligning with our previous work confirming superior CO₂ clearance across oronasal mask valves. Crucially, the plateau valve—despite its controlled-leak mechanism—showed no rebreathing reduction at clinically relevant settings, this provides actionable guidance for optimizing NIV trial protocols and clinical implementation. Furthermore, valve-dependent performance variations revealed some imperatives: rigorous monitoring of ventilator parameters (trigger sensitivity, control, and ventilation monitoring) during exhalation valves selection; development of ventilators with real-time intentional leak monitoring to optimize ventilator performance; and future studies also need to validate these findings across different patient populations (e.g., restrictive lung diseases) through bench-to-bedside investigations.

Several methodological limitations warrant consideration. This in vitro analysis employed a simulated COPD model with fixed hypercapnia (PetCO₂ 80 mmHg) and restricted ventilator pressure settings, thereby failing to capture physiological variations in respiratory patterns, metabolic CO₂ fluctuations, or compensatory mechanisms in chronic respiratory failure. The restriction to 12/4 and 16/4 cmH₂O pressure settings—while clinically relevant—precludes evaluation across the full therapeutic spectrum, and the absence of human variables (e.g., airway humidity, spontaneous breathing variability) limits clinical extrapolation. Though providing rigorously controlled conditions for quantifying CO₂ rebreathing dynamics, these constraints necessitate: (1) expanded pathophysiological models (e.g., restrictive disease); (2) systematic valve evaluation across incremental pressures (e.g., IPAP 8–25 cmH₂O; EPAP 4–12 cmH₂O); and (3) prioritized clinical validation measuring ventilator performance parameters, and gas exchange efficacy in target populations.

Our bench study demonstrated that different exhalation valves significantly affect CO₂ rebreathing and ventilator performance during NIV. CO₂ rebreathing was significantly lower with the mask valve than with the single-arch, whisper swivel, and plateau valves during NIV. Except for the V60 ventilator, ventilator performance parameters of the plateau valve differed from those of the other valves; additionally, the rate of difference in V_T between the ventilator and the simulated lung exceeded 10% for all exhalation valves. Further clinical studies are needed to validate these results in vivo, taking into account patient variability in respiratory mechanics, ventilator settings, and CO₂ production.