which ventilators and modes can be used to deliver noninvasive

Which Ventilators and Modes Can Be Usedto Deliver Noninvasive Ventilation?

Robert L Chatburn RRT-NPS FAARC

IntroductionGoals and Objectives of Ventilation: Invasive Versus NoninvasiveWhich Modes Best Meet the Goals of NIV?Evaluating Ventilator Performance

Lung ModelsPerformance Variables

Volume Control Versus Pressure ControlCMV Versus IMV Versus CSVPressure Support Versus Proportional AssistWhich Ventilators Best Provide the Modes?

Noninvasive VentilatorsHome-Care VentilatorsICU VentilatorsPerformance Comparisons

How to Choose the Right TechnologyKey Decisions

Summary

A wide variety of ventilators and modes can be used to deliver noninvasive ventilation (NIV). Tonavigate successfully through the many options, the clinician must first have a clear understandingof the goals of mechanical ventilation: namely, safety, comfort, and timely liberation. Examining thespecific objectives associated with these goals, we can distinguish priorities for NIV. This paperreviews the methods of achieving those objectives by reviewing the characteristics of ventilationmodes and how those characteristics are measured in performance-evaluation studies. This reviewprovides the basis for a simple procedure for selecting the most appropriate NIV technology for thepatient and the environment of care. Key words: mechanical ventilation, ventilator, ventilation mode,noninvasive ventilation, NIV, bi-level, proportional-assist ventilation, pressure support, lung model.[Respir Care 2009;54(1):85–99. © 2009 Daedalus Enterprises]

Robert L Chatburn RRT-NPS FAARC is affiliated with the Section ofRespiratory Care, Respiratory Institute, Cleveland Clinic, and with theDepartment of Medicine, Lerner College of Medicine, Case WesternReserve University, Cleveland, Ohio, and with Research and ClinicalServices, Strategic Dynamics, Scottsdale, Arizona.

Mr Chatburn has a relationship with Cardinal Health. He reports no otherconflicts of interest related to the content of this paper.

Mr Chatburn presented a version of this paper at the 42nd RESPIRATORY

CARE Journal Conference, “Noninvasive Ventilation in Acute Care: Con-troversies and Emerging Concepts,” held March 7-9, 2008, in Cancun,Mexico.

Correspondence: Robert L Chatburn RRT-NPS FAARC, RespiratoryTherapy, M-56, Cleveland Clinic, 9500 Euclid Avenue, Cleveland OH44195. E-mail: [email protected].

RESPIRATORY CARE • JANUARY 2009 VOL 54 NO 1 85

Introduction

You have assessed the patient. He seems to meet manyof the standard textbook criteria1,2 for considering nonin-vasive ventilation (NIV) (Table 1). But now you face abewildering array of technological decisions. The selec-tion of patient-ventilator interface is not too difficult (ie,nasal mask, nasal pillows, oronasal mask, full-face mask,or helmet), except that manufacturers called oronasal masks“full face” until someone invented a real full-face (“totalface”) mask3 and now there is a nomenclature issue. Butwhat about the ventilator? Which ventilator do you use,and which mode do you select? And how do you evenknow which mode is which when the literature is full ofconfusing terms and acronyms, such as volume-cycled,pressure-cycled, volume-limited, control mode, assist-con-trol (A/C) mode, pressure-control ventilation (PCV), in-termittent positive-pressure ventilation (IPPV), bi-levelpositive airway pressure (BiPAP), spontaneous/timed (S/T)mode, volume-assist/control, pressure-assist/control, pres-sure-support ventilation (PSV), and proportional-assist ven-tilation (PAV)?

In this paper I will take a step-by-step approach to thisproblem, by first reviewing the general goals of mechan-ical ventilation and how they might be emphasized differ-ently in NIV versus invasive ventilation (ie, intubated pa-

tients). Next I will review the basic nomenclature andclassification scheme for ventilation modes, to dispel someof the confusion about jargon. Having identified whichmodes best meet the goals of NIV, I will then discusswhich ventilators best provide the modes. Finally, I willconclude with a simplified algorithm for selecting themost appropriate ventilator and mode for NIV in a givensituation.

Goals and Objectives of Ventilation:Invasive Versus Noninvasive

The goals of mechanical ventilation, in the most basicterms, can be thought of as (1) do no harm, (2) promotepatient comfort, and (3) liberate the patient from me-chanical ventilation as soon as possible (Table 2). Nodoubt these goals are the same whether treating thepatient invasively or noninvasively. However, if we lookat the specific objectives associated with these goals,perhaps we can distinguish priorities for the 2 approachesto ventilation. For example, intubated patients are usu-ally more critically ill, so safety objectives (eg, follow-ing blood gas values to assure adequate ventilation andusing a lung-protective ventilation strategy) tend to pre-dominate over comfort objectives (eg, intubated patientsare often sedated). Table 3 lists some criteria to deter-mine if safety should be the priority. If none of those

Table 1. Criteria for Considering Noninvasive Ventilation

Disease processes and situations that definitely indicate NIV (strongevidence1,2)

COPD exacerbationFacilitation of weaning/extubation in patients with COPDAcute cardiogenic pulmonary edemaImmunocompromised (hematologic malignancy, bone marrow or

solid-organ transplant, AIDS)

Inclusion criteria (at least 2 should be present)Use of accessory musclesParadoxical breathingRespiratory rate � 25 breaths/minDyspnea (moderate-to-severe or increased in COPD)PaCO2

� 45 mm Hg with pH � 7.35PaO2

/FIO2� 200 mm Hg

Exclusion criteriaApneaHemodynamic or cardiac instabilityUncooperative patientFacial burn or traumaHigh risk of aspirationCopious secretionsAnatomic abnormalities that interfere with gas delivery

COPD � chronic obstructive pulmonary diseaseAIDS � acquired immune deficiency syndromeFIO2 � fraction of inspired oxygen

Table 2. Goals and Objectives of Mechanical Ventilation

Do no harm (minimize risks and maximize safety)Provide adequate gas exchange by optimizing the ventilation-

perfusion ratioAcceptable blood gas valuesMinimal dead spaceAdequate cardiac output and blood pressure

Avoid ventilator-induced lung injury by optimizing the lungvolume/pressure ratio (optimize mechanics)

Optimal PEEPOptimal tidal volume

Minimize risk of ventilator-associated pneumoniaPromote patient comfort

Avoid patient-ventilator asynchrony by optimizing the ratio ofventilator work output to patient work output, ie, ventilatorydemandOptimal trigger and cycle sensitivities (spontaneous breaths)Optimal inspiratory time and peak inspiratory flow (mandatory

breaths)Liberate patient from ventilator as soon as possible

Minimize duration of mechanical ventilationOptimal criteria for weaning screen, spontaneous breathing trial,

and extubationOptimal adherence to protocols and avoidance of delays

PEEP � positive end-expiratory pressure

WHICH VENTILATORS AND MODES CAN BE USED TO DELIVER NONINVASIVE VENTILATION?

86 RESPIRATORY CARE • JANUARY 2009 VOL 54 NO 1

criteria are met, and there are no other countervailingcircumstances, the priority should be comfort. Comfortobjectives tend to predominate with patients treated withNIV, because these patients are usually awake, alert,and desire to interact with the world in as normal afashion as possible, particularly if the NIV is for achronic or permanent condition. If this distinction be-tween invasive ventilation and NIV is valid, then itshould guide our selection of ventilation modes.

For a brief overview of how ventilation modes are de-scribed and classified, see the Appendix, which provides thebasic concepts and terminology for the rest of this paper.

Which Modes Best Meet the Goals of NIV?

Having defined a “mode” as a basic ventilatory pattern(for the simplified purposes of this paper), we can nowreview the literature with minimum confusion regardingterminology. In other words, I will translate the variousnames used in the cited articles into the standardized termsfor ventilatory patterns (with the exception that I will alsorefer to continuous positive airway pressure [CPAP], PSV,and PAV).

But, before evaluating the merits of various modes, weneed to review the laboratory evaluation of ventilator per-formance in general. The key difference between invasiveventilation and NIV is that NIV typically involves air leaks,because NIV uses a mask instead of an endotracheal tube.NIV air leaks can be large enough to affect the ventilator’striggering and cycling functions and the rate of inspiratorypressure rise. Thus, the standard laboratory setup for as-sessing ventilator performance under NIV conditions is toconnect the ventilator to a simulated load (ie, lung modelwith variable resistance and compliance) with a circuit thathas a variable (ie, controllable) leak.

Evaluating Ventilator Performance

Lung Models

The most common lung model for this type of experi-mental setup is a double-bellows model, such as the Train-ing and Test Lung (Michigan Instruments, Grand Rapids,Michigan).4 With this device, one bellows is connected toa “driving” ventilator set to simulate the patient’s ventila-tory muscle actions. The other bellows is connected to theventilator being evaluated. The connection to the test ven-tilator is often via a simulated patient head fitted with amask. The connection between the mask and the test ven-tilator includes a variable leak (eg, a valve that opens tothe atmosphere). Flow transducers are placed in the circuitbetween the mask and the test ventilator and between thelung simulator and the mask. Pressure, volume, and flowsignals are then digitally recorded and analyzed with acomputer. Figure 1 shows a typical setup.5

The problem with the above-mentioned setup is that thelung simulator mechanical properties (ie, resistance andcompliance) are limited to a finite number of discrete spring-tension settings or calibrated flow resistors. In addition,researchers often have to create their own data-analysisalgorithms (eg, to calculate trigger delay). Fortunately,there is a better test lung commercially available. In myopinion, the best available lung simulator is the ASL 5000(IngMar Medical, Pittsburgh, Pennsylvania).6 It is basedon a computer-controlled piston that moves inside a cyl-inder. Volume (V) as a function of time (t) is measureddirectly as piston displacement. Flow (V) is the derivativeof volume with respect to time (dV/dt). Airway pressure(P) is governed by the equation of motion for the respira-tory system:

P � �1/C�V�t� � R�dV/dt� � Pmus�t�

where C is compliance, R is resistance, and Pmus(t) isventilatory muscle pressure as a function of time, whichis either set to zero (ie, passive inflation and deflation) oris defined by the operator to simulate active breathing.The simulator can be programmed as a single-compart-ment or double-compartment model (ie, 1 or 2 lungs).Compliance is simulated by moving the piston accordingto dV � (C)(dP). The relationship between pressure andvolume can be made nonlinear to better approximate themodel to the S-shaped pressure/volume response curve ofa real patient. Resistance is defined by dP � (R)(dV/dt), sothe piston is moved at a speed of dV/dt � dP/R. Differentvalues for resistance can be selected for flows in the di-rection of inspiration and expiration, and the resistor set-tings can be linear or parabolic. Parabolic resistors havebeen the choice for most physical resistors, because im-plementations of linear resistors demand linear flow over

Table 3. Criteria to Determine if a Noninvasive VentilationCandidate’s Condition Warrants Safety as a Priority

Apnea riskCentral apneaFailure to trigger due to intermittent leaks

Hypoventilation RiskSevere COPD exacerbationFailed extubation

Hypoxemia RiskHigh FIO2

requirementHigh PEEP requirement

COPD � chronic obstructive pulmonary diseaseFIO2 � fraction of inspired oxygenPEEP � positive end-expiratory pressure



the whole expected flow range. The ASL 5000 simulatoravoids those difficulties and provides a response that rep-resents both types.

Spontaneous breathing effort is simulated by specifyinga Pmus function. The operator can set the frequency, am-plitude, and various parameters of the waveform. An op-tional device, the Simulator Bypass and Leak Valve Mod-ule (IngMar Medical, Pittsburgh, Pennsylvania), provides3 levels of simulated leak (Fig. 2), to provide the repro-ducible leaks required for ventilator-performance testingwith simulated NIV scenarios. Figure 3 shows a typicalexperimental setup.7 The ASL 5000 comes with softwarespecifically designed to test ventilator performance, so ithas a wide variety of waveform-analysis features and cal-culated variables.

In this article I report results obtained with an ASL 5000,programmed with the settings shown in Table 4.

Performance Variables

Patient-ventilator asynchrony during NIV can be clas-sified according to the 4 breath phases: the switch fromexpiration to inspiration; inspiration; the switch from in-spiration to expiration; expiration.

Leaks in the delivery system cause problems in Phase 1,such as delayed triggering, decreased sensitivity (missedtrigger efforts), or increased sensitivity (auto-triggering).Phase 2 effects include a decreased rate of inspiratorypressure rise. Phase 3 effects occur if the leak is largeenough to prevent inspiratory flow decay to the cycle thresh-old. In Phase 4, leaks may cause loss of positive end-expiratory pressure (PEEP).

Some of the variables that have been defined (eg, by theASL 5000) to evaluate ventilator performance (Fig. 4)include:

Drop to Pmin: The change in airway pressure from PEEPto the minimum airway pressure (ie, pressure below base-line before inspiratory flow starts going in the positivedirection) caused by a patient inspiratory effort. Some au-thors have referred to this as “trigger pressure.” However,the precise meaning of trigger pressure is the pressurethreshold the ventilator uses to start inspiratory flow. ThePmin recorded at the airway opening during the triggerphase of a breath is the result of several factors, includingthe inspiratory effort, the patient circuit compliance, andthe ventilator’s gas-flow delivery characteristics, so theairway pressure typically continues downward even thoughthe ventilator has started inspiratory flow. Thus, Pmin maybe less than the true trigger pressure. Indeed, “trigger pres-sure” is not even relevant when inspiration is flow-trig-gered.

Inspiratory time (TI): The time from the start of inspira-tory effort (ie, Pmus beginning to go below zero) to the startof expiratory flow (or to maximum volume of theASL 5000).

Time to Pmin: The time from the start of inspiratoryeffort (ie, Pmus beginning to go below zero) to the mini-mum airway pressure.

Time from Pmin to baseline (PEEP): The time it takesfor airway pressure to go from minimum airway pressureto baseline (PEEP).

Trigger response time: The time from the start of in-spiratory effort to the point where airway pressure reachesbaseline (PEEP) (ie, the sum of Time to Pmin and Timefrom Pmin to baseline).

Patient trigger work: The integral of Pmus with respectto volume from the start of inspiratory effort to the pointwhere airway pressure returns to PEEP after Pmin.

Pressure-time product: The integral of airway pressure(relative to PEEP) with respect to time from the start of

Fig. 1. Bellows lung simulator setup to evaluate ventilator performance during noninvasive ventilation. (Adapted from Reference 5, withpermission.)



inspiratory effort to the point where airway pressure re-turns to PEEP after Pmin.

Inspiratory T90: Time for airway pressure to rise to 90%of the steady-state value. This is an index of pressure-waveform distortion due to a leak. A lower value indicatesthat pressure rises slower because of flow loss from theleak.

Cycling delay: TI minus inspiratory effort time (ie, thetime that Pmus is below zero). A positive number meansinspiratory flow from the ventilator ended after the effortended (delayed cycling), and a negative number meansthat inspiratory flow ended before the effort ended (pre-mature cycling).

Volume Control Versus Pressure Control

The first consideration in evaluating modes is whetherthere is a difference between volume control (VC) andpressure control (PC) when trying to ventilate in the pres-ence of large and possibly unstable leaks. Rather than tryto guess or infer the answer from experimental data, wecan much more reliably model the situation mathemati-cally. For this purpose we borrow a schematic and analy-

ses from electrical engineering. Figure 5 shows the respi-ratory system represented as an electrical resistanceconnected in series with a capacitance. The leak is mod-eled as a resistance in parallel with the respiratory system.Pressure control is modeled as a step change in voltageapplied across the model, and volume control is modeledas a step change in electrical current. The mathematicalmodel that corresponds to this electrical model is analo-gous to the equation of motion. The equation is solved forvolume (as a function of time) using the Laplace Trans-form (Fig. 6).8 In Figure 6 the intersection of the 2 solidlines represents the target tidal volume (VT) delivered witha preset TI and the no-leak condition. The shape of the 2sets of curves show that pressure control delivers more ofthe target VT for any time less than the target TI than doesvolume control. However, the difference in leak betweenthe 2 modes (ie, the vertical distance between “leak off”and “leak on” in Fig. 6) becomes clinically unimportant asTI decreases (eg, as breathing frequency increases). Thetheoretical waveforms in Figure 6 show that a leak causeslung volume to rise exponentially rather than linearly involume control. In effect, volume control has been changedto pressure control, because the pressure drop across theleak acts like a pressure-limiting valve. What I concludefrom this analysis is that, theoretically, pressure control isbetter than volume control.

Another theoretical advantage of pressure control is that,for the same VT, pressure control results in a lower peakinspiratory pressure9 (provided that the TI is longer thanabout 3 time constants of the respiratory system). This istrue when volume control is achieved with a constant flow.A lower peak inspiratory pressure may allow less masktightness and better patient comfort. However, some ven-tilators (eg, the Aequitron LP-10, Aequitron Medical/Mallinckrodt, Plymouth, Minnesota) deliver volume-con-trolled breaths with a descending-ramp flow waveformthat approximates the shape of the flow waveform duringpressure control, so the peak inspiratory pressure is similar(Fig. 7).

But are the differences important in the real world ofNIV? To answer that question I compared VC-CMV toPC-CMV with an Aequitron LP-10 ventilator. I achievedpressure control by setting the VT to 1.0 L and using themechanical pressure-limit knob to adjust peak inspiratorypressure. I connected the ventilator to the ASL 5000 lungsimulator, as described above, and set the ventilator todeliver a VT of about 480 mL. I recorded delivered VT atbaseline (no leak) and at leak settings of 1, 2, and 3 on theSimulator Bypass and Leak Valve Module. Figure 8 showsthe results. The VT with volume control dropped 32% (atleak setting 3), compared to only 6% with pressure con-trol. This confirms not only that the theoretical results arevalid but also that the effect may be clinically important.The take-home message is that pressure control should

Fig. 2. Pressure versus flow characteristics of the Simulator By-pass and Leak Valve Module used with the Ingmar ASL 5000 lungsimulator. (Courtesy of Ingmar Medical.)



provide a more stable ventilatory support than volumecontrol if there is a large and variable leak.

There are, however, advantages to volume control. Pis-ton-driven portable volume-control ventilators can providehigher ventilating pressure, which may be required in pa-tients with high respiratory-system impedance (eg, due toobesity). These ventilators can operate considerably longeron battery power than blower-driven pressure control de-vices. Finally, patients using volume-control ventilatorscan be taught to “stack” (ie, double-trigger) breaths toattain a high inspired lung volume to increase cough flow.2

Some studies have compared pressure control to volumecontrol in terms of physiologic variables. Two studies10,11

found no difference in oxygenation or ventilation betweenthe two. Girault et al12 found that volume control hadlower inspiratory work load than did pressure control. Vi-tacca et al13 found no difference in NIV success rate be-

tween volume control and pressure control. Shonhoferet al14 found that pressure control was successful in themajority of patients after an initial treatment with volumecontrol. Two patients in that study had intractable flatu-lence with volume control but were adequately treatedwith pressure control (presumably due to a lower peakinspiratory pressure). One third of the patients who ini-tially did well on volume control failed on pressure con-trol, despite the fact that the 2 modes were matched forrate and VT during the day. The authors speculated thatone reason for not responding to pressure control may bethe postural change of lung and chest wall mechanics dur-ing nocturnal ventilation, which leads to hypoventilation.

Fig. 3. Experimental setup to evaluate ventilator performance dur-ing noninvasive ventilation with the Ingmar ASL 5000 test lung andSimulator Bypass and Leak Valve Module. (Courtesy Robert LKacmarek PhD RRT FAARC.)

Fig. 4. Flow, pressure, volume, and muscle pressure waveforms toillustrate the definitions of ventilator-performance variables. Verti-cal axis is pressure (� 10), volume, and flow (� 10). Horizontal axisis time. Pmin � change in airway pressure from PEEP to the min-imum airway pressure. PEEP � positive end-expiratory pressure.T90 � time for airway pressure to rise to 90% of the steady-statevalue.

Fig. 5. Electrical circuit that models a leak in parallel with therespiratory system. R � resistance. Tube � endotracheal tube.V � flow. C � compliance. P � pressure. (From Reference 8.)

Table 4. Settings Used With the Ingmar ASL 5000 Lung Simulator

Single-compartment model to simulate a patient with COPDMechanical Variables

Uncompensated residual volume: 0.5 LInspiratory resistance: 10 cmH2O/L/sExpiratory resistance: 10 cmH2O/L/sCompliance: 80 mL/cmH2O

Pmus VariablesFrequency: 15 breaths/minAmplitude: 5 cmH2OInspiration: 5% (of ventilatory period)Hold: 3%Release: 15%Pmus modified by airway pressure (backing off): 50%Baseline mask leak: 4 L/min at mean airway pressure of 8 cmH2O



The main advantage of volume control is a more stableVT with changing lung mechanics. The main advantagesof pressure control are a more stable volume delivery withleak and better patient-ventilator synchrony because pres-sure control allows the patient to adjust inspiratory flow.Adaptive pressure control15 attempts to achieve both setsof advantages by making inspiration pressure-controlledbut letting the ventilator automatically adjust the peak in-spiratory pressure to achieve a preset target VT. To datethere have been no studies of NIV with adaptive pressurecontrol, nor have there been any ventilators that use adap-tive pressure control for their NIV modes. However, theShonhofer et al14 study suggests that adaptive pressurecontrol may have avoided the failure of those patients whohyperventilated at night with simple pressure control. Onthe other hand, a substantial leak might confound the adap-tive control algorithm. More study of adaptive pressurecontrol for NIV is warranted.

CMV Versus IMV Versus CSV

If, in general, comfort objectives tend to predominatewith patients treated with NIV, then we can argue thatCSV is better than CMV, because with CSV the patientcan modify both the timing and size of the breath, whichincreases synchrony of breathing efforts with ventilatorgas delivery and thus promotes comfort. With CMV, boththe size of the breath and the TI are arbitrarily set by theclinician. Excellent patient assessment in the moment canassure good synchrony with CMV, but the problem is thatthe operator is not always there to readjust the settings ifthe patient desires to change his or her breathing pattern.On the other hand, if safety concerns predominate, thenCMV is a better option.

I can think of no argument (nor have I found any liter-ature) to justify the use of IMV in the traditional sense,meaning the delivery of volume-controlled mandatorybreaths along with pressure-controlled spontaneous breaths.That seems to me the worst of both worlds: possibly asyn-chronous mandatory breaths interspersed with variable-volume spontaneous breaths, which would result in a pat-tern of highly variable VT delivery and work of breathing(ie, from breath to breath).

Fortunately, the IMV provided by specialized noninva-sive ventilators is different from that provided by home-care and intensive care unit (ICU) ventilators. The defini-tion of IMV is that spontaneous breaths can be deliveredbetween mandatory breaths.4 Traditionally, IMV was im-plemented on ventilators as a preset backup rate for man-datory breaths, and they were delivered regardless of thespontaneous breathing activity. However, with noninva-sive ventilators there may be 2 kinds of IMV. For exam-ple, with the Respironics BiPAP S/T ventilator, in the“timed” mode, IMV is delivered traditionally (ie, manda-tory breaths are time-triggered, time-cycled, and deliveredat the set rate, independent of the patient’s spontaneousbreathing efforts). However, in the “spontaneous/timed”mode, mandatory breaths (ie, time-triggered but flow-cy-cled) are delivered only if the spontaneous breath rate fallsbelow the set IMV rate. This has clinical importance be-cause the operator may be tempted to increase the rate (aswith traditional IMV) for a patient who is hypoventilating(ie, rapid shallow breathing pattern), only to find that noth-ing happens, because the set rate is still below the patient’strigger rate. What is needed in that case is to increase thepressure-support level.

The need for routine use of a backup rate has not beenestablished,16 although if patient safety is the priority, itmakes sense to use CMV or IMV. For example, in patientswith neuromuscular disease, setting the backup rate slightly

Fig. 6. Delivered volume as a function of time in pressure-con-trolled ventilation (PCV) or flow-controlled ventilation (FCV), whichis synonymous with volume control. (From Reference 8.)

Fig. 7. Characteristic waveforms with the Aequitron LP10 ventila-tor during volume-control ventilation. Vertical axis is pressure(� 10), volume, and flow (� 10). Horizontal axis is time.



below the normal breath rate while the patient is awakeshould provide maximum respiratory muscle rest duringsleep.17

The take-home message is that if the main objective iscomfort, use CSV; otherwise, use CMV on a home-careventilator, or IMV on a noninvasive ventilator or ICUventilator that has an NIV option.

Pressure Support Versus Proportional Assist

If we pursue the idea of using CSV to promote comfort,we naturally ask how patient-ventilator synchrony can beenhanced. Much has been written about the possibility ofasynchrony with PSV.18 As with other modes, trigger de-lay can increase the patient’s work of breathing. The pres-surization rate can be too slow (which causes the patient’sfeeling of not getting enough flow) or too fast (whichcauses the patient to reflexively activate expiratory mus-cles and prematurely terminate the breath). If the pressure-support level is set too low, the patient may not get a largeenough VT. This may also lead to an TI shorter than thepatient’s neurological inspiratory drive time, which willcause double-triggering breaths and auto-PEEP. On theother hand, if the pressure-support level is set too high, thepatient may feel overinflated and, again, actively exhale toterminate the breath, which increases the work of breath-ing. Setting the pressure-support level too high also tendsto lengthen the TI, because the peak inspiratory flow ishigher, so the flow cycle threshold (which is usually set asa percent of peak flow) takes longer to reach, and, again,may urge the patient to actively terminate the breath. Thiseffect may also be seen in patients with long respiratorysystem time constants. Finally, if the leak is large enough,inspiratory flow may never decay to the cycle threshold,

and, again, this forces the patient to actively terminate thebreath, or to wait until the time-cycle backup mechanismis activated. Newer ventilators allow the operator to adjustthe pressure rise time and the cycle threshold so that, withdiligence and attention to both the patient and the venti-lator graphics, PSV can be fine-tuned to be synchronous inmost situations. But you can see the problem: a skilledoperator with the time to do the fine-tuning is a scareresource in most settings. Advanced PSV algorithms havebeen developed,19 and I would argue that the more fine-tuning the ventilator can do automatically, the better. Weneed to build more intelligence into ventilators, becausewe cannot guarantee it at the bedside.

PAV offers a theoretical advantage over PSV in that theinspiratory pressure and the inspiratory cycle threshold areboth under more control of the patient and do not needbreath-by-breath fine-tuning by an operator. Indeed, therehave been numerous comparisons of PSV and PAV in theliterature, with conflicting results. Compared to PSV, PAVhas been found to reduce indices of inspiratory muscleeffort,20,21 or not,22 to be associated with more rapid im-provements in physiologic variables,23 or not,24 and is seenas more comfortable.22,23 Neurally adjusted ventilatory sup-port seems promising for the same reasons as PAV, but weneed more research to verify that assumption.

The take-home message seems to be that if you havePAV, use it in preference to PSV. Unfortunately, the onlyventilator available in the United States with an approvedPAV mode explicitly recommends that it not be used forNIV, because, in its present incarnation, PAV� does nothave leak compensation and therefore cannot properly eval-uate lung mechanics with a mask.

In summary, if comfort is the primary objective, usePAV when available; if not, use PSV, making sure to

Fig. 8. Effect of leak size on delivered tidal volume. VC-CMV � volume-controlled continuous mandatory ventilation; PC-CMV � pressure-controlled continuous mandatory ventilation.



fine-tune pressure rise time, peak inspiratory pressure, andcycle threshold to assure synchrony. If safety is the pri-mary objective, use PC-CMV or PC-IMV. In the rare casethat the patient needs cough assist (and can be taught theskill), use VC-CMV.

Which Ventilators Best Provide the Modes?

There are 3 general classes of ventilator that can be usedfor NIV: devices specially designed for the task (I will callthem “noninvasive ventilators”), standard home-care ven-tilators, and ICU ventilators.

Noninvasive Ventilators

Noninvasive ventilators are pressure controllers, andtheir “claim to fame” is that they are designed to performwell with relatively large system leaks. Noninvasive ven-tilators provide either PSV or PC-IMV. However, as de-scribed above, regardless of the backup rate, mandatorybreaths are only delivered if the spontaneous breath ratefalls below the set rate.

Simple Noninvasive Ventilators. Simple noninvasiveventilators (sometimes called “bi-level” or “BiPAP”) ven-tilators are inexpensive, small, and relatively easy to op-erate. A simple noninvasive ventilator requires an oxygenbleed-in modification to control the fraction of inspiredoxygen (FIO2

). Unlike other types of ventilators, a simplenoninvasive ventilator does not use an exhalation valve.Instead, it regulates a constant flow of air up or down asneeded to control airway pressure. As such, a simple non-invasive ventilator uses a single-limb patient circuit, andmost such circuits have a fixed leak near the mask, toprevent rebreathing carbon dioxide. A simple noninvasiveventilator is typically powered by a blower, which con-sumes a relatively large amount of power, so few simplenoninvasive ventilators operate on batteries (those that do,not for long). A simple noninvasive ventilator is morepractical than an advanced noninvasive ventilator for homeuse, and simple noninvasive ventilators are also popular inhospitals. The operator interface is usually limited to asmall liquid-crystal display screen that shows text mes-sages only. Simple noninvasive ventilators offer pressure-support and CPAP, and a few have a ramp setting thatgradually increases the CPAP level so the patient can moreeasily fall asleep (ie, for treating obstructive sleep apnea).Most simple noninvasive ventilators have no alarms, ex-cept perhaps an alert if the mask leak is too great. Figure 9shows representative pressure, volume, and flow curveswith BiPAP (with maximum leak setting 3 on the Simu-lator Bypass and Leak Valve Module).

Advanced Noninvasive Ventilators. Advanced nonin-vasive ventilators (eg, Respironics Vision and HamiltonRaphael) offer control over FIO2

, graphic monitoring, andalarms. For instance, the alarms on the Respironics Visioninclude high-pressure limit, low-pressure limit, apnea, highrate, low rate, and low minute ventilation. Advanced non-invasive ventilators are most often used in hospitals, inboth intensive and acute care areas. Figure 10 shows rep-resentative waveforms from a Respironics Vision.

Home-Care Ventilators

Small, piston-driven ventilators (eg, Lifecare PLV-100and Aequitron LP-10) have been the standard home-careventilator for decades. Such ventilators are well suited forpatients who need continuous ventilatory support and thosewith severe chest-wall deformity or obesity, who needhigh inflation pressure.17 However, home-care ventilatorsare heavier and more expensive than simple noninvasiveventilators. Home-care ventilators are typically used forVC-CMV or VC-IMV, but some can be set to deliverpressure-control ventilation. These devices are easy to use,offer adequate alarms, and can operate on battery power,but require an oxygen bleed-in modification to controlFIO2

. The major concern about home-care ventilators forNIV is that they are not designed to compensate for leaks,which may cause triggering problems. Figure 7 showscharacteristic waveforms from an Aequitron LP10.

ICU Ventilators

ICU ventilators were originally designed to ventilateintubated patients, with minimal or no leak. Although someearly ventilators had ingenious leak-compensation devices,especially for infants with uncuffed tubes (eg, BournesLS104–150), leaks have generally caused triggering prob-lems with ICU ventilators. With the growing popularity ofNIV, some newer ICU ventilators (eg, Maquet Servo-i,Hamilton G5, Drager Evita XL, Newport e500, ViasysVela, Respironics Esprit) have “noninvasive modes” alongwith their standard inventory of ventilatory patterns. Theseare really not unique ventilatory patterns; they are justPC-IMV, but when activated, the ventilator automates leakcompensation (eg, adjusts bias flow and trigger sensitivity)and deactivates some nuisance alarms. The operator inter-faces look different from other modes; the intent is tomimic the settings on noninvasive ventilators. The advan-tage of using an ICU ventilator is that you have all thesophisticated control and monitoring features you couldwant. The disadvantage, of course, is the higher cost andcomplexity. Figure 11 shows characteristic waveforms froma Hamilton G5.

Recently, Vignaux et al5 evaluated the performanceof ICU ventilators with noninvasive modes. They con-



cluded that leaks interfere with several key functions ofICU ventilators. There was auto-triggering, decreasedpressurization rate, and delayed cycling. Activating theNIV modes on these ventilators corrected the auto-trig-gering on all the ventilators they tested, but pressuriza-tion and delayed cycling were corrected on only someof the ventilators. Vignaux et al noted a wide variationamong the tested ventilators in the efficiency with whichtheir NIV modes handled the various dysfunctions causedby leaks.

More recently, Ferreira et al7 conducted a similarstudy and found that at a baseline leak (that simulatedproper mask fit) all the tested ventilators were able todeliver adequate VT and maintain airway pressure andsynchrony with the (simulated) patient. As the leak wasincreased, all the ventilators (except the Maquet Servo-iand the Respironics Vision) needed manual adjustmentof sensitivity or cycling threshold to maintain adequateventilation.

Performance Comparisons

I have collected representative performance data for the3 types of ventilators used for NIV (see Table 5). Theexperiments were conducted with the Ingmar ASL 5000lung simulator, as described above. Data were collectedand analyzed with the ASL 5000’s data-collection soft-ware, version 2.2. Only mean values (for a minimum of 10breaths) are displayed, because the coefficients of varia-tion were negligible for all variables. I tested leak condi-tions of no leak, baseline, and settings 1, 2, and 3 on theSimulator Bypass and Leak Valve Module. However, forsimplicity, I report only the data for leak setting 3 (max-imum).

With maximum leak, all the ventilators were able todeliver a VT within 6% of the no-leak condition. As ex-pected, the pressure drop during triggering was least forthe dedicated noninvasive ventilators (Respironics BiPAPand Vision). Surprisingly, the G5 had the shortest trigger-response time, although it was the only ventilator thatrequired manual adjustment of the sensitivity to preventauto-triggering. Except for the (easily corrected) auto-trig-gering by the G5, all the ventilators operated without prob-lems with all leaks.

I would like to call attention to one subtle aspect ofventilator-performance testing that I don’t think has beendiscussed in the literature. As noted above, the IngmarASL 5000 can be set so that Pmus is modified by the airwaypressure. The intent (according to the ASL 5000’s operatormanual) is to simulate the situation where a patient makesless effort during inspiration as the airway pressure rises.Of course, whether or not patients actually react that wayis debatable. With this model, differences in airway pres-sure rise time will create differences in the Pmus profile.The problem is that there is a great deal of variability in

Fig. 9. Characteristic waveforms with bi-level positive airway pres-sure (BiPAP). Vertical axis is pressure (� 10), volume, and flow(� 10). Horizontal axis is time.

Fig. 10. Characteristic waveforms with the Respironics Vision ven-tilator. Vertical axis is pressure (� 10), volume, and flow (� 10).Horizontal axis is time.

Fig. 11. Characteristic waveforms with the Hamilton G5 ventilator.Vertical axis is pressure (� 10), volume, and flow (� 10). Horizontalaxis is time.



the airway pressure waveforms of different ventilators.This means that any attempt to measure performance vari-ables that depend on Pmus (eg, pressure-time product orimposed work of breathing) will suffer the confoundingeffect of having different simulated patient efforts. I be-lieve this is a limitation of the data in Table 5, and there-fore advise caution in the interpretation of that data. If Iwere to design a formal study for comparison of ventila-tors’ performance characteristics, I would have done itdifferently. In short, if the intent of a study were, forexample, to compare the work of triggering of variousventilators (and hence different pressure delivery patterns),then the patient effort should be kept constant (ie, deacti-vate the Pmus modified by Paw or “backing off” feature ofthe Ingmar ASL 5000).

How to Choose the Right Technology

We lack robust evidence for choosing ventilators andmodes for most (if not all) NIV applications in the acute-care setting. Therefore, we must rely on reasoning fromfirst principles. Having reviewed the theory and evidenceon modes and ventilators, we can establish a decision frame-work for matching the appropriate technology to a givenpatient and care environment. Figure 12 shows an algo-rithm than can be used as a general guide.

Key Decisions

The decision nodes (diamonds) in Figure 12 are num-bered. The questions are:

1. Does the patient’s condition prompt us to be moreconcerned about safety or comfort (see Table 3)?

2. If safety is the priority, the next question is whetherthe patient is in the hospital. If so, generally, more sophis-ticated ventilators will (or should) be available.

3. Is the patient in the ICU or emergency department? Ifso, the best option is to initiate PC-IMV with an advancedICU ventilator that has an NIV option. These devices typ-ically only provide IMV, and a backup rate is desirablebecause the priority is safety. An advanced ICU ventilatorprovides the flexibility to switch to or from NIV as re-quired, without needing additional machines. If the patientis not in the ICU or emergency department, an advancedventilator should still be provided, but the more practical(ie, less costly and easier to use) alternative to an ICUventilator is a dedicated (advanced) noninvasive ventila-tor.

4. If the patient is not in the hospital, would thepatient benefit from the cough assistance afforded bybreath-stacking? If so, the only way to provide breath-stacking is with volume control on a home-care venti-lator. (Airway clearance should not be an issue thatguides ventilator selection in the hospital environment,because there are many alternative therapies and per-sonnel to provide them.)

5. Advanced ventilators (in a hospital environment)can probably provide all the power necessary to venti-late any patient. In the home-care environment, how-ever, patients with severe restriction (eg, obese patientsand those with severe chest deformities) may need ahigher ventilating pressure than can be provided by asimple noninvasive ventilator. There may be exceptions,of course, so you should check the available ventilators’specifications.

6. In a home-care environment, mobility may be animportant issue, so consider whether a battery-poweredventilator is available. Newer, blower-type noninvasiveventilators may provide battery operation for a limitedtime (eg, about an hour). Again, check the manufacturer’sspecifications and remember that battery age can affectoperating time.

Table 5. Performance Characteristics of Representative Ventilators Used for Noninvasive Ventilation*

Ventilator LeakTidal Volume

(mL)Leak Volume

(mL)Drop to Pmin

(cm H2O)Time to Pmin

(ms)

Time fromPmin to PEEP

(ms)

Trigger ResponseTime(ms)

Respironics BiPAP Pro None 322 0 –0.3 83 75 158Maximum 320 2 –0.3 67 167 234

Respironics Vision None 334 0 –0.6 88 66 154Maximum 323 11 –0.7 95 63 158

Aequitron LP 10 None 485 0 –1.1 146 39 185Maximum 454 31 –0.9 156 30 186

Hamilton Galileo None 441 0 –0.8 117 18 135Maximum 464 –23 –0.9 102 19 121

* See text for definitions of performance characteristics.



Summary

A wide variety of ventilators and modes can be used todeliver NIV. To navigate successfully through the manyoptions, the clinician must first have a clear understandingof the goals of NIV, particularly whether safety or comfortis most important factor for the patient in question. Pres-sure control is generally better able to compensate forpatient/interface leaks than is volume control. Spontane-ous-breathing modes such as pressure support and propor-tional-assist ventilation probably provide the best patientcomfort.

REFERENCES

1. Kacmarek RM. Noninvasive positive pressure ventilation. In: WilkinsRL, Stoller JK, Kacmarek RM, editors. Egan’s fundamentals of re-spiratory care. 9th edition. St. Louis: Mosby Elsevier; 2009:1091-1114.

2. Hill N. Noninvasive positive-pressure ventilation. In: Tobin MJ. Prin-ciples and practice of mechanical ventilation. 2nd edition. New York:McGraw-Hill; 2006:433-471.

3. Roy B, Cordova FC, Travaline JM, D’Alonzo GE Jr, Criner GJ. Fullface mask for noninvasive positive-pressure ventilation in patientswith acute respiratory failure. J Am Osteopath Assoc 2007;107(4):148-156.

4. Michigan Instruments. The evolution of mechanical cardiopulmo-nary resuscitation (CPR) through the decades. http://www.michiganinstruments.com. Accessed November 3, 2008.

5. Vignaux L, Tassaux D, Jolliet P. Performance of noninvasive ven-tilation modes on ICU ventilators during pressure support: a benchmodel study. Intensive Care Med 2007;33(8):1444-1451.

6. IngMar Medical. Take flight with respiratory simulation: IngMarMedical breathing simulators, lung models, and test lungs. http://www.ingmarmed.com. Accessed November 3, 2008.

7. Ferreira JC, Chipman DW, Kacmarek RM. Trigger performance ofmid-level ICU mechanical ventilators during assisted ventilation: abench study. Intensive Care Med 2008;34(9):1669-1675.

8. Chatburn RL, Volsko TA, El-Khatib M. The effect of airway leak ontidal volume during pressure- or flow-controlled ventilation of theneonate: A model study. Respir Care 1996;41(8):728-735.

9. Chatburn RL. Fundamentals of mechanical ventilation. ClevelandHeights: Mandu Press; 2003.

10. Navalesi P, Fanfulla F, Frigerio P, Gregoretti C, Nava S. Physiologicevaluation of noninvasive mechanical ventilation delivered with three

Fig. 12. Noninvasive ventilation (NIV) mode-selection algorithm for clinicians unfamiliar with NIV. SNIV � simple noninvasive ventilator.ANIV � advanced noninvasive ventilator. ICU � intensive care unit. ED � emergency department. AICV � advanced intensive careventilator. HCV � home-care ventilator. PC-CMV � pressure-controlled continuous mandatory ventilation. VC-CMV � volume-controlledcontinuous mandatory ventilation. PC-IMV � pressure-controlled intermittent mandatory ventilation. PAV � proportional-assist ventilation.PSV � pressure-support ventilation.



types of masks in patients with chronic hypercapnic respiratory fail-ure. Crit Care Med 2000;28(6):1785-1790.

11. Meecham Jones DJ, Paul EA, Grahame-Clarke, Wedzicha JA. Nasalventilation in acute exacerbations of chronic obstructive pulmonarydisease: effect of ventilator mode on arterial blood gas tensions.Thorax 1994;49(12):1222-1224.

12. Girault C, Richard J, Chevron V, Tamion F, Pasquis P, Leroy J, Bon-marchand G. Comparative physiologic effects of noninvasive assist-control and pressure support ventilation in acute hypercapnic respiratoryfailure. Chest 1997;111(6):1639-1648.

13. Vitacca M, Rubini F, Foglio K, Scalvini S, Nava S, Ambrosino N.Noninvasive modalities of positive pressure ventilation improve theoutcome of acute exacerbations in COLD patients. Intensive CareMed 1993;19(8):450-455.

14. Schonhofer B, Sonneborn M, Haidl P, Bohrer H, Kohler D. Com-parison of two different modes for noninvasive mechanical ventila-tion in chronic respiratory failure: volume versus pressure controlleddevice. Eur Respir J 1997;10(1):184-191.

15. Branson RD, Chatburn RL. Controversies in the critical care setting.Should adaptive pressure control modes be utilized for virtually allpatients receiving mechanical ventilation? Respir Care 2007;52(4):478-485.

16. Mehta S, Hill NS. Noninvasive ventilation. Am J Respir Crit CareMed 2001;163(2):540-577.

17. Hill NS, Eveloff SE, Carlisle CC, Goff SG. Efficacy of nocturnal

nasal ventilation in patients with restrictive thoracic disease. Am RevRespir Dis 1992;145(2 Pt 1):365-371.

18. Brochard L, Lellouche F. Pressure-support ventilation. In: Tobin MJ.Principles and practice of mechanical ventilation. 2nd edition. NewYork: McGraw-Hill; 2006:221-250.

19. Du HL, Amato MB, Yamada Y. Automation of expiratory trigger sen-sitivity in pressure support ventilation. Respir Care Clin North Am2001;7(3):503-517.

20. Grasso S, Puntillo F, Mascia L, Ancona G, Fiore T, Bruno F, et al.Compensation for increase in respiratory workload during mechanicalventilation. Am J Respir Crit Care Med 2000;161(3 Pt 1):819-826.

21. Wysocki M, Meshaka P, Richard J, Similowski T. Proportional-assist ventilation compared with pressure-support ventilation duringexercise in volunteers with external thoracic restriction. Crit CareMed 2004;32(2):409-414.

22. Varelmann D, Wrigge H, Zinserling J, Muders T, Hering R, PutensenC. Proportional-assist versus pressure support ventilation in patientswith acute respiratory failure: cardiorespiratory responses to artificiallyincreased ventilatory demand. Crit Care Med 2005;33(9):1968-1975.

23. Gay PC, Hess DR, Hill NS. Noninvasive proportional assist venti-lation for acute respiratory insufficiency. Am J Respir Crit Care Med2001;164(9):1606-1611.

24. Fernandez-Vivas M, Caturla-Such J, Gonzalez de la Rosa J, Acosta-Escribano J, Alvarez-Sanchez B, Canovas-Robles J. Noninvasivepressure support versus proportional assist ventilation in acute re-spiratory failure. Intensive Care Med 2003;29(7):1126-1133.



Discussion

Nava: I want to point out the differ-ence between synchrony and interac-tion. People frequently call synchronya problem when it’s not, and I wonderif you agree with me. Synchrony comesfrom chronos, which is Greek for time,and is therefore a matching of the 2phases: the start of inspiration and theend of inspiration. Interaction is re-lated to timing but is more “gross”(eg, ineffective efforts, double-triggering, auto-triggering), right? Theideal ventilator should fit these two,but synchrony is especially difficultto measure, because even transdia-phragmatic pressure is not goodenough; you would need electromyo-graphy. Interaction is another thing.Interaction is about auto-triggering, in-effective efforts, and double-trigger-ing, which is very important. The onlyventilation modality I know that looksat chronos is neurally activated ven-tilation. Which is more important: syn-chrony or interaction?

Chatburn: I am not sure I wouldagree with your distinction betweensynchrony and interaction. Auto-trig-gering, double-triggering, and ineffec-tive efforts can be also defined in termsof time or chronos.

We need to work on the nomencla-ture of ventilator-patient interactionand sort out these details. Our view ofreality has a lot to do with how wedefine words. The way I look at ven-tilator-patient interaction and syn-chrony versus asynchrony is that it re-lates to the 4 phases of a breath: theswitch from exhalation to inhalation;the inspiratory phase; the switch frominspiration to expiration; and the ex-piratory phase. Things can go wrongin any of the phases, and most of themhave to do with time.

In my bench studies, in trying tomake a consistent volume pattern withthe different ventilation modes I foundthat it often comes down to adjustingthe inspiratory time: it’s chronos. Youwant the mechanical inspiratory time

to be as consistent as possible with theneural inspiratory time, but there’s alsofactors such as how much PEEP youapply in the inspiratory phase so thatyou can balance the intrinsic PEEP ifthere is any, and how high should theinspiratory pressure be? If it’s too highor low, it’s uncomfortable. It’s not justtiming; it has to do with the size of thebreath as well.

Nava: Do you think it’s importantto quantify or estimate leaks? If youwant to quantify leak during NIV, youneed to measure both inspired and ex-pired VT.

Chatburn: Well, hopefully the ven-tilator is smart enough to manageleaks, and that seems to be the trendin ventilator design.

Gay: I don’t think the portabilityand battery power of those ventilatorsis an issue, because we can put a bat-tery on anything we want. But what Ireally want to talk about is your NIValgorithm. I think it’s difficult to thinkthis through because our prioritieschange continuously with virtually allof our NIV patients. One of the mostfundamental priorities intervenessometimes when these patients getcomfortable enough to fall asleep, andthen how do they behave? The syn-chronicity of their ventilatory prob-lem often becomes apparent, sowhether there is an important sleep/ventilation issue needs to be in thealgorithm.

Chatburn: Do you think it’s do-able? To have a decision map like that,to give people some general way togo?

Gay: I think we need to address thechronic component of hypoventilationsomewhere in there, because it’s a dif-ferent mindset as you start to thinkthrough these treatment models.

Kacmarek: My concern about youralgorithm is that it is totally based onventilation mode, and a given venti-lation mode may function differentlyfrom one ventilator to another. Thething to emphasize more is the venti-lator’s capability to provide NIV. Ev-ery one of the ventilators on the mar-ket that has an NIV mode functionsdifferently, and their ability to handleleaks and ensure synchrony is vastlydifferent. So it would be wrong to as-sume that one ICU ventilator is equiv-alent to another for NIV. One venti-lator may do a wonderful job with leak,whereas another one cannot compen-sate for any leak. So before we canuse an algorithm like that, there has tobe similarity in the operation of theNIV modes on ICU ventilators.

Chatburn: Absolutely true. Thepublished study and your data empha-sized that. Hopefully we’ll get to somestandardization of performance so weare comparing apples to apples.

Hill: I like the idea of an algorithm,but the devil is always in the details.How are you going to prioritize, forexample, safety versus comfort? Whatabout gas exchange? For example,some bi-level ventilators lack an ox-ygen blender and do a lousy job andmay even be dangerous for hypoxemicpatients. That should be in the algo-rithm. The available evidence is defi-cient on this topic. For instance, I don’tthink we have the evidence to say thatan advanced critical care ventilator ismore or less effective than an advancednoninvasive ventilator for different pa-tients.

I also think it’s a little premature tohave cough assist in the algorithm, atleast for acute care. In home care Ithink there’s a lot more evidence tosupport it, but in the acute care set-ting, we really don’t have any evi-dence other than from patients withneuromuscular disease. It’s very im-portant to start thinking about how tostructure an algorithm, but we have tobe very careful to specify where we



have evidence and where we don’t.We have to be clear on our prioritiesand be prepared to change the algo-rithm as they change.

Also, when you get down to com-paring the details of specific adjust-ments to flow and timing and so forth,you can make certain modes performvery much like other modes and getaround a lot of the problems related todifferent performance characteristics.Should the algorithm consider thattoo? And are we going to see propor-tional-assist ventilation any time soon?

Chatburn: My comment about thatis that you know too much. If youknow enough about mechanical ven-tilation, you can ventilate anybodywith anything. This algorithm is morefor the uninitiated—somebody whoknows something about ventilators butdoesn’t have much experience withNIV and doesn’t know how to fine-tune NIV. There’s got to be some high-level approach to put people on theright path, and then they can get ex-perience and learn how to do the fine-tuning. That’s just the way I look at it.Maybe that’s impractical. I don’tknow.

Hess: I’m glad you brought up theissue of power and portability. A prac-tical problem in the hospital is that ifwe start NIV in the emergency depart-ment or out in the ward and then weneed to transfer the patient to the ICU,

sometimes interrupting NIV for just afew minutes results in the patient de-teriorating. There are noninvasive ven-tilators that have internal batteries, andthe other possibility is a medical gradeuninterruptible power supply.

Chatburn: How long does it last?

Hess: You get 30 minutes to an hourfor sure, and considerably more de-pending on the settings.

Kacmarek: If it’s working right, it’ssupposed to last 8 hours.

Kallet: I’m glad you brought up thepoint about getting too fancy with thelung models. You can’t mimic whatthe patient is going to do unless youhave a neural network. I think withthe IngMar ASL 5000 lung model youcan vary the resistance and compli-ance, but you can’t model a pleuralpressure gradient. In other words,there’s no sheet of scarred or edema-tous lung tissue that blunts the trans-mission of airway pressure. If there’snothing blunting that positive airwaypressure when you’re mimicking neg-ative muscle pressure, then the venti-lator’s performance looks much dif-ferent than you’d see with someonewho has ARDS [acute respiratory dis-tress syndrome]. When you can’t eas-ily transmit that positive airway pres-sure, you can’t outpace the respiratorymuscles. If you can’t displace the chest

with the ventilator faster than the pa-tient’s muscles can, the patient willcontinue to perform a lot of work. Itneeds to be stressed that these modelshave limits. I think lung-model testingcan deceive us as to how well theseventilators actually perform.

Chatburn: Those are excellentpoints. The value of a physical lungsimulator or a mathematical model isthat you get a basic understanding ofwhat should happen in an ideal world.That is your base of reference, andthen you go out and see what happensin the real world, and it’s degraded tosome extent, but it’s a lot easier to gofrom theory of what should be—thepristine point of view—to the realworld than to try to infer back theother way.

Kacmarek: The thing that’s im-possible to ensure is that a patientmaintains exactly the same ventila-tory pattern as 10 ventilators are ap-plied. A patient’s ventilatory patternchanges moment to moment. In ad-dition, an IRB [institutional reviewboard] would not approve a study inwhich there is a risk but no potentialbenefit to the patient. Lung-modelstudies must be taken for what they.They simply evaluate the function-ing of ventilators under the same,very well defined, consistent set ofcircumstances.



which ventilators and modes can be used to deliver noninvasive

Documents