Airway Pressure Release Ventilation (APRV) for ARDS

Up until Question 3 from the cursed first paper of 2023, APRV had never come up in the CICM exams, and one might have been forgiven for expecting that was unlikely to ever appear. The reason for this is that APRV is not a well-accepted rescue technique, nor widely practised, nor especially well-researched; and so one might think it would be inappropriate to include something like that in the exams. You can't claim that a question on APRV should be able to fail or pass a future intensivist on the grounds that this is an essential part of their core curriculum. Still, it is used around the world as a rescue therapy for the most severe ARDS, and so it seems it should form a part of the intensivist' toolbox, as to exclude it on the basis of there being no evidence would also exclude a vast swath of other ventilation strategies and adjuncts (hello, heliox and inhaled prostacyclin).

For the reader with a few spare minutes, the amazingly detailed paper by Andrews et al (2022) is the strongest and most coherent recent piece on APRV, and should be essential reading for anyone preparing to defend their use of APRV in a clinical environment full of sceptics. Excellent resources from PulmCrit and EMCrit are also available, including peer-reviewed original works by erudite contributors. The time-poor candidate should skip directly to the pros and cons section of this LITFL summary, and read no further. 

The openmost of the open lungs

Working on the premise that driving pressure (the change in pressure between PEEP and inspiratory pressure) does all the damage in ARDS by means of cyclic atelectasis, one might come to the conclusion that breaths in general are harmful, and that tidal breathing in general should be abolished. Pressure should be constant and high so that all the alveoli are maximally recruited and none collapse at any stage during the respiratory cycle because there is no cycle. For the open lung enthusiast, the ideal ventilator waveform is this:

Now, most reasonable people would agree that this is in fact apnoea, and that some kind of ventilation needs to still occur in order for gas exchange to take place (as CO₂ clearance with this model would obviously be suboptimal). The other constant-pressure-non-breathing mode (high frequency oscillation) achieves this by vibrating the gas column inside the lung, encouraging some gas mixing and therefore the clearance of CO₂:

However, we can all agree that this mode is quite dead. Most oscillator hardware around the world now lies dormant under coats of dust. The reason for this is probably the failure of this mode to protect the patient from VILI by increasing the power of ventilation (i.e. imparting too much energy to lung tissue) and by being so difficult to tolerate that heavyhanded sedation and paralysis were required, making more it difficult to recover from critical illness. These factors and others have conspired to make HFOV look really bad in clinical trials, such that OSCAR showed no benefit and OSCILLATE was stopped early because of harm. Surely, one might say, there must be some mode that walks a line of compromise between the need to sustain a high mean airway pressure while allowing some CO₂ removal to take place, in some patient-acceptable way that requires less sedation?

Yes, colleagues, that mode is APRV.  

Observe: the mean airway pressure is constant, fixed at a safe value, and lung-pleasingly high. At the same time an occasional pressure release occurs, allowing a small amount of gas to escape from the patient, permitting the clearance of CO₂. Moreover, all this happens without any violent lung-bashing oscillation, which means sedation can be weaned and even more CO₂ clearance could occur because the patient is permitted to take breaths throughout the period of high pressure:

Sounds brilliant. So thought Stock et al (1987), who first described this idea and demonstrated its effects using an animal model. That original article was light on praise, in the sense that the authors did not tout the advantages of this method as much as they could have, but subsequent works have filled in the blanks.

Advantages of APRV

They are said to be numerous:

• If you like open lung, you can't lung it more open than that. Or, to rephrase this in a more professional way, this ventilation strategy prioritises alveolar recruitment and minimises the risk of shear stress and cyclic atelectasis, therefore theoretically reducing ARDS-related biotrauma from ventilation. 
• AutoPEEP keeps the alveoli recruited: gas trapping is actually central to how APRV works, whereas in conventional ventilation it is merely a desirable side-effect of high respiratory rates required for low-tidal-volume ventilation.
• Breaths are few. The power of ventilation with APRV should be only quite modest, which should theoretically reduce the amount of total energy transferred to the lung tissue, contributing to improved outcomes by minimising ventilator-associated lung injury.
• Tidal volumes are smaller, or at least fewer, and the lower minute ventilation should be lung-protective, again using decreased power of ventilation as the main argument (i.e. fewer joules transferred to the lung tissue per unit time)
• Paralysis and deep sedation are not essential. The patients are in fact encouraged to breathe spontaneously, as the mode gradually transitions to CPAP while they recover. With their neuromuscular junctions unmolested, these patients probably have a better chance of weaning off the ventilator (Goligher et al, 2017).
• Basal aeration should be better. The diaphragms are doing some useful work, and the negative pleural pressure they generate in the bases of lung should theoretically recruit the bases of the lung. 
• Haemodynamics may be better. Spontaneous respiratory effort should improve venous return and ameliorate some of the ill effects of the high pressures one ends up using with this mode of ventilation
• Secretion clearance may improve.  This seems to get repeated a lot. It appears many of the proponents of APRV believe that the high expiratory flow rate during the breath release will produce some improved clearance of lower airway secretions by increasing the shear force of the escaping gas on the sputum, analogous to what a cough does. This appears to be supported by at least one study in which the investigators ventilated preserved pig lungs and observed the movement of faux secretions along a 7.0 endotracheal tube. In case you are wondering, the authors chose to add green food colouring to their simulated sputum for no clear scientific reason. This pseudosnot was a lot more mobile in the APRV group as compared to the conventional ARDSNET protocol.

Disadvantages of APRV

• Barotrauma. Though the data from ARDS publications does not seem to support this, anecdotally there is a higher risk of pneumothorax and pneumomediastinum with this strategy, perhaps partly due to the fact that these patients are awake and therefore free to cough violently while on a PEEP of 30. Clearly an abrupt pressure challenge to an already overstretched lung with infection-softened parenchyma would be unwelcome. In practice, the small trials and case series describing APRV do not report a vastly excessive rate of chest drain insertion - for example, 4% (two patients) in Lim et al (2016).
• The patients need to have a lot of fluid volume, which is contrary to the usual dry lung strategy pursued for ARDS patients (FACTT trial, NEJM, 2006). The main reason for this is usually the haemodynamic effect of a sustained high intrathoracic pressure, which markedly decreases preload. Both of the ways of dealing with this are unsatisfactory (more vasopressors, or more fluid).
• It's unfamiliar. Ok, this is not really a very strong disadvantage, but when pushed against the wall most reasonable people would agree that, for reasons of your safety,  the intensivist looking after you should be familiar with your mode of ventilation, and comfortable with troubleshooting it.  Not to mention the swaths of junior doctors and nurses who need to troubleshoot the mode when you are in REM sleep in the middle of the night. To say that APRV has yet to reach the right of levels of widespread acceptance would be a hilarious understatement of the superstion fear and jealousy expressed by mainstream authors. No other mode of ventilation has ever been referred to as "the devil's spawn" in a professional publication, by a panel of experts who admitted having minimal experience with APRV shortly before they dismissed it as worthless.
• RV strain is thought to be a greater risk, considering the pressures involved. The effect on RV afterload would probably exclude some patients from APRV, such as those with existing severe pulmonary hypertension, or those who have the propensity to develop right-to-left intracardiac shunting.
• Tidal volumes are usually not lung protective. With a P_low of 0 cm H₂O, the release breaths can often be confrontingly huge, which would appear to violate the norms of ARDS ventilation. 
• Derecruitment or shear stress can result from the release breaths. For a split second, the pressure in the respiratory circuit is allowed to fall to a very low value, and some argue that this could lead to derecruitment and atelectasis.
• Without spontaneous breathing, this is just inverse ratio ventilation. A lot of the benefits are lost if the patient requires a lot of sedation for some other non-ARDS reason.

A practical approach to APRV

There are four main variables that need to be adjusted, which are sufficiently different from the variables used to control conventional ventilation that it would be valuable to go through them in some detail.

• T_low: the time spent at the low pressure (P_low)
• P_low: the release pressure
• P_high: the constant high pressure
• T _high: time spent at the constant high pressure

Or, in picture form, in case that somehow helps:

T_low is the most important setting

Because it determines the main factor that could make or break the APRV strategy, which is whether the patient will derecruit or not. When starting the patient on APRV, this setting is usually titrated to minimise derecruitment. It is usually very short: a long expiratory time would allow derecruitment to occur and would lose a lot of the desirable autoPEEP that keeps the alveoli open. The starting setting could be 0.3-0.5 seconds, depending on how long it takes to get to 75% of peak expiratory flow.

Why 75%? This is not an arbitrary number. It is chosen as a compromise between satisfactory gas release and the potential for alveolar collapse. It appears we have been doing this for longer than we have had solid data to support it. When most people from the noughties quoted this specific percentage, their reference seems to be a 2005 article by Nader Habashi, whose recommendation for T_low was mainly based on porcine models of acute lung injury by Markstaller et al (2001) and Neumann et al (1998). These investigators observed the changes in alveolar aeration by performing high resolution CT scans of ventilated animals in varying stages of ARDS, and their data suggested that derecruitment occured over an average timeframe of around 0.6-0.8 seconds, but neither had looked at expiratory flow in any sort of detail. Habashi's expiratory flow rate suggestion (to aim for 50-75% of peak expiratory flow) seems to have come from Guttmann et al (1995) who measured the relationship of flow to the change in expiratory volumes in ARDS patients.  Modern papers and resources instead reproduce the images and data from Kollisch-Singule et al (2016 and 2014) who looked at alveolar recruitment along a range of PEFR values and determined the optimal PEFR percentage by alveolar microscopy. When peak expiratory flow was limited to 75%, the difference between inspiratory and expiratory alveolar volumes was only about 10%, the most eloquent support of this specific strategy.

So: T_low determines whether your lung-protective APRV strategy is actually lung-protective, which will determine its success or failure in the long term. It may be necessary to fiddle with this variable as the patient's respiratory physiology changes, i.e. where the time constants of the lung units change and the expiratory flow rate increases or decreases. As a broad generalisation:

• Extremely stiff lungs in early severe ARDS tend to have very rapid early expiratory flow rates, and shorter T_low values (0.4 or even 0.3 seconds) may be required. If the patient has extremely poor compliance and this variable is not adjusted to a PEFR of 75%, the lung-protective benefits of APRV may be lost (the PEFR may end up being below 50%, which means many of the alveoli will de-recruit)
• As the patient starts getting better the elastic energy stored in the lungs decreases and the T_low can be stretched to 0.5 seconds or longer. If this is not adjusted, the expiratory volume may end up being small, and increasing autoPEEP may develop. The PEFR percentage will be observed to creep up. This may have absolutely no disadvantage, as by this stage the lung pathology and gas exchange may be improved to the point where CO₂ will not accumulate even with this extra gas trapping effect, but most reasonable people would agree that it is not optimal.

P_low determines the pressure gradient for release flow

This is the pressure drop during the pressure release breath. It is difficult to judge what the best setting here should be, but most people would agree that this should be minimal, and ideally either zero, or close to zero (eg. 5 cm H₂O in Zhou et al, 2017). The objective is to provide the largest gradient for expiratory flow, so that the maximum amount of CO₂-rich gas can take advantage of this precious short opportunity for expiration. Though one might have set a P_low of zero, practically that pressure is not achieved in the circuit because expiratory flow maintains some positive pressure in the airway, and the actual "PEEP" is in fact slightly higher.  That relationship is well demonstrated in these excellent images of APRV captured in the wild by Deranged Physiology contributor Dr George Zhou, a different Zhou to the one who published the trial:

How much volume could you possible exhale during this split second? A substantial amount. From the in vivo ventilator recordings above we can see that the peak expiratory flow was close to 70L/min. If we crudely approximate the mean expiratory flow rate to be closer to 50L/min, that gives us an expired volume of around 416 ml during that half-second pressure release, which is close enough to a lung-protective volume for a 70kg patient. With twelve such releases over a minute, one could achieve a respectable minute volume of approximately 5L - plus whatever else happens via spontaneous breaths. 

On occasion, it becomes necessary to adjust the P_low, particularly where hypoxia is a major problem: 

• If oxygenation remains poor on otherwise satisfactory settings, one could increase the P_low. This would produce more autoPEEP and reduce the risk of derecruitment. The tidal volumes would also decrease, and become even more lung-protective. This is the preferred move in response to worsening hypoxia, as opposed to increasing the P_high.
• If the CO₂ removal was poor for some reason, theoretically one could consider dropping this value to zero, if it were not zero already (but most often it is so low that this adjustment would have minimal effect)

P_high maintains the safety of the APRV strategy

And also the oxygenation, as this is the sustained pressure that recruits the alveoli. And the driving pressure for the release breaths (it's not all P_low , it takes two to gradient).  The P_high should be set to something close to the patient's peak inspiratory pressure while on a conventional pressure-controlled mode, and ideally below 30 cmH₂O. This means that at the initiation of APRV this variable is almost always set to a high value, something like 28 cmH₂O, because the patient has severe ARDS and their plateau pressure is inevitably high. It should not be set much higher than that, as this is the safe maximum above which lungs tend to get damaged (with the exception of the morbidly obese patient). As the patient improves P_high should be slowly reduced.

T_high determines much of the CO₂ clearance

CO₂ clearance is also controlled by the amount of expiratory flow afforded by the P_low and by the duration of that flow (i.e. the T_low). Unfortunately the first of those variables is usually fixed at zero, and the other one is too important to mess with (because it influences the oxygenation and lung-protectiveness of the overall strategy). This means that T_high ends up being the remaining variable you can adjust to increase the minute volume.

A normal strategy when starting APRV is to set the T_high somewhere around 9 times longer than the T_low, i.e. a 9:1 I:E ratio (for example, for a T_low of 0.5 seconds, the T_high would be 4.5 seconds, with a total breath cycle time of 5 seconds). That gives twelve cycles per minute, and usually achieves a minute volume that is at least not worse than those seen in conventional ventilation for the same patient. If the result produces a high CO₂, the T_high can be reduced to give more breaths per minute, thereby generating a larger minute volume.

T_high is also the variable used to wean the patient once they start getting better. This is generally described as the "drop and stretch" approach, where the T_high is gradually increased, and the P_high is gradually decreased, with the patient breathing spontaneously through the whole process, so the mode slowly and surely transitions to CPAP. 

With this mode overall being the province of expert opinion rather than evidence, it is not surprising that everyone has their own recipe for the weaning process. Authorities such as Josh Farkas and Emily Parent from EMCrit suggest to drop the P_high by 2cm H₂O at every increment, and to prolong the T_high by 0.5-2 seconds, leaving the other variables basically untouched. This should produce a situation where the patient, awake and comfortable, is breathing spontaneously on CPAP with a PEEP of whatever the last P_high was (Josh suggests 15-20), and this is interrupted twice every minute by an increasingly irrelevant release breath. From here it is not a huge leap to conventional ventilation or even extubation.