Bench testing plays a crucial role in respiratory medicine for evaluating non-invasive ventilation (NIV) devices. It allows detailed assessment of ventilator performance, patient–ventilator interaction, and physiological responses without the risks of patient trials. With NIV increasingly used across patient groups and ventilator technology evolving, bench testing has become more important.

Simulation devices for bench testing range from simple passive lung simulators like SmartLung® and QuickLung® that are affordable but limited in replicating spontaneous breathing, to mid-level models offering some adjustment of breathing effort (i.e., Michigan testLung®). Advanced active simulators like the ASL 5000 (IngMar Medical) and newer models such as TestChest® (Neosys) and Alveo® (IMT Analytics) provide complex and realistic respiratory mechanics, accurately mimicking muscle forces for detailed device evaluation. Table 1 summarizes current available simulators.

Educational manikins add anatomical realism and allow for testing interfaces and upper airway behaviour. Recent trends include 3D-printed head models based on actual patient imaging for more realistic simulations [1].

These testing systems can simulate a variety of breathing patterns including normal, obstructive, and restrictive profiles. They allow control over parameters like air leakage, respiratory rate, and pressure support, enabling detailed ventilator benchmarking. Setups may include external instrumentation to measure airflow, pressures, leaks, etc, or rely in internal sensors of the advanced lung simulators.

Despite many bench studies since the early 2000s, protocols and outcome measures still vary between labs [2]. While research has expanded to include ventilator comparisons, interface testing, and clinical scenarios, more consistency in methods is needed for stronger clinical application.

When physicians prescribed a drug, they assume that the drug will have the efficacy demonstrated in clinical trials. They know common side effects and can explain to their patients why this drug will have such effect on their disease. But drugs are drugs, they have their chemically active compound that is always the same and their absorption body need to be stable. Indeed, drugs need to have stable pharmacokinetics. To illustrate that impact, switching from immediate release to long-term release drugs requires a medical prescription. For drugs, the only variability comes from the patient (who can have one cytochrome or another, who can have other drugs that creates interactions…). One may argue that NIV is the same: the active chemical compound is just air that is pressured into the lungs. It may be true. But then, the way that the drug is delivered, i.e. the pharmacokinetics of NIV should be stable. As each NIV manufacturer has developed its own turbine, its own algorithm, each NIV has a different pharmacokinetics. Hence, in the field of NIV, when we decide to administer a drug named NIV, we have variability coming from the patient (as any regular drugs) but we have another source of variability different NIV-pharmacokinetics. Things get even more complicated when no data is needed to issue a NIV ventilator on the market and make it ready to use in patients. Therefore, physicians have many drugs (NIV-ventilators) with different pharmacokinetics (that they do not know in most cases). As NIV is delivered in frails patients, as NIV is delivered to a small number of patients (as compared to sleep apnoea), and as patients requiring NIV have different respiratory mechanics, it is impossible to assess all NIV-pharmacokinetics in humans. However, knowing pharmacokinetics is essential, the only segue is to have data from bench tests. Bench testing excels in delivering safety and efficiency, enabling respiratory researchers and device manufacturers to examine ventilator responses to any scenarios, leaks, and asynchrony without jeopardizing patient safety. The laboratory environment facilitates rapid data gathering and robust comparative analysis at a fraction of the cost and logistical complexity of patient trials. Researchers gain deeper physiological insight, with precise control and measurement of parameters such as compliance, resistance, and flow that would be challenging to manipulate bedside. Standardized setups allow objective device comparisons, fostering innovation and raising industry standards. Regulatory bodies and clinicians alike benefit from quantitative, reproducible metrics, supporting evidence-based choices and safer clinical deployments.

Bench testing has already shown its clinical relevance during the COVID pandemic and other crisis in the field of NIV [3–5]. It may be argued that it does not replace clinical evaluation in humans and that is true. However, it is also a necessary step to ensure adequate behavior of the ventilator. Indeed, bench tests should be considered as a pre-clinical model to assess ventilators. This is how the industry is developing their ventilators. Nonetheless, these pre-clinical assessments need to be open and repeated in independent laboratories to ensure data quality. To bridge the gap with real-life assessments, bench test also needs to progress and validate their findings using real-life data [6–8].

Bench testing of non-invasive ventilation (NIV) devices is important but limited by significant variability in methods, making comparisons challenging. Different lung simulators vary in performance, and experimental setups differ in circuit configuration, environmental conditions (e.g., ATPS vs. BTPS), and signal measurement techniques. Analytical methods also lack standardization, with metrics like PTAP, PTP300, and PTP500 varying in calculation methods and components included, causing inconsistent reporting [2]. Statistical analyses differ, for example, in comparing median values versus individual breaths, impacting interpretation of ventilator performance. Bench models often fail to replicate complex patient physiology accurately, especially in simulating air leaks that are typically represented by simple valves (“proportional” leaks) rather than reflecting non-proportional, realistic leaks which affect ventilator synchrony [9]. Attempts to mimic real-life conditions face trade-offs between technical accuracy and physiological realism, limiting the predictive value of bench testing. The linearity of conventional lung models represents perhaps the most significant limitation of current bench testing approaches. Moreover, single or dual compartment models may not reflect the heterogenicity of sick lungs. Thus, the response of the ventilator may vary in the real patient in a non-predictable way. Even more, changes in compliance at different lung capacities may be challenging to be reflected by current simulators [10]. Only the more advanced, digitally controlled simulators may offer some capabilities to modify the pressure–volume relationship (hysteresis and volume-dependent compliance) as reflected in Fig. 1. Also, resistance may be challenging when considering it in a linear or in a parabolic way.

Fig. 1.

Pressure–volume curve tuning capabilities in three lung simulators. Interfaces for the most advanced lung simulators, showing capabilities of nonlinear compliance. (A) ASL5000® (IngMar Medical®) allows for settings of the whole compliance, and upper and lower inflection point, and differentiates inspiratory and expiratory resistance. (B) Alveo (IMT Analytics) in a similar way allows for upper and lower inflection points. (C) TestChest offers a more complex parameter settings regarding compliance, but only allows for global resistance setting.

Upper airway dynamics during NIV remain poorly replicated in bench setups. The interaction between positive pressure and compliant upper airway structures – particularly during sleep – significantly impacts NIV efficacy. Most bench models employ rigid conduits that fail to mimic physiological airway collapse or the complex nasopharyngeal flow dynamics. Glottic reflex is rarely simulated due to technical challenges. This limitation is especially relevant when evaluating triggers and cycling, as upper airway resistance substantially alters work of breathing and patient–ventilator synchrony. Studies employing Starling resistors to mimic upper airway collapse fail to explain laryngeal reactivity [11] and there is significant heterogeneity in resistor design across studies.

When evaluating technologies such as forced oscillation, which is incorporated in some ventilators and AutoPAP devices to automatically adjust expiratory pressure or CPAP levels, it should be acknowledged that the resulting responses may not be physiological [1]. This occurs because the underlying algorithms are designed to assess the “absorption” of pressure pulses by lung tissue rather than their reflection from a rigid surface, such as a lung simulator piston. Moreover, these discrepancies may be further amplified when comparing simulators based on different mechanical principles, such as rubber bellows, piston-driven, or other designs, as algorithms are designed to evaluate the “absorption” of pressure pulses by lung tissue, rather than impacting in a metal surface such as the lung simulator piston [12]. The differences between simulators introduce uncertainty regarding the reproducibility of published results (Table 1, current features of marketed simulators).

Finally, most bench studies evaluate ventilators under conditions that poorly represent the clinical scenarios for which the devices are intended. Few studies incorporate simulated patient efforts that accurately reflect the respiratory drive, timing, and variability seen in acute respiratory failure. While there has been some work in the field of appraising realistic effort patterns [13], most of the trials employ pure sinusoidal patterns (in best case when they mentioned it [14]).

The disconnect between bench testing and patient outcomes represents perhaps the most critical limitation of this approach. While bench studies can identify technical differences between devices, they cannot predict how these differences will translate into clinical efficacy. Usually patients under NIV are not sedated, may be awake or sleep, anxious or relaxed, with muscular impairment or fighting the ventilator, and successful NIV depends not only on technical performance but also on patient comfort, tolerance, and cooperation-factors that bench testing cannot adequately assess.

While bench testing provides valuable preliminary data on device performance, its findings must be interpreted cautiously within the context of these substantial limitations. The path from bench to bedside remains fraught with complexities that simplified mechanical models cannot address, necessitating robust clinical validation of any insights derived from bench testing methodologies. Some papers where this bench-to-bedside is developed can serve as a way to go for future trends [7,8].

The frontier of simulation is rapidly advancing. High-flow therapy is being integrated into bench platforms to reflect clinical trends and complex humidity-flow-oxygen dynamics [15]. Upper airway simulation, leveraging 3D-printing and fluid dynamics, better captures mask fit and dead space effects, such as rebreathing. Mathematical models and computational simulations incorporate real-world variability and non-linear respiratory mechanics. Digital twins promise patient-specific virtual testing, potentially bridging the gap between technical evaluation and individualized therapy. AI-driven analysis, biomimetic modeling, and smart sensor integration herald a new era of automated, physiologically relevant, and comprehensive simulation, though with their own demands for validation and standardization. Virtual reality and IoT training environments will embed bench realism in frontline education, forging stronger feedback loops between simulation and patient experience.

AI tools (DeepL and Copilot) have been used to improve the English writing of the manuscript, but no generative AI was used without supervision to produce content.

Javier Sayas Catalán reports grants from Menarini and Philips, consultancy fees from Philips Respironics and ResMed.- Honoraria for lectures from Philips Respironics, support for attending meetings from Nippon Gases.

Maxime Patout reports grants from Fisher & Paykel, ResMed and Asten Santé, consultancy fees from Philips Respironics, Air Liquide Medical, ResMed, Asten Santé and GSK, payment or honoraria for lectures, presentations, manuscript writing or educational events from Philips Respironics, Asten Santé, ResMed, Air Liquide Medical, SOS Oxygène, Antadir, Chiesi, Jazz Pharmaceutical, Loewenstein, Fisher & Paykel, Bastide Medical, Orkyn and Elivie, support for attending meetings from Asten Santé and Vitalaire, participation on a data safety monitoring board or advisory board with ResMed, Philips Respironics and Asten Santé, stock (or stock options) with Kernel Biomedical, and receipt of equipment, materials, drugs, medical writing, gifts or other services from Philips Respironics, ResMed and Fisher & Paykel.