A young woman presented to Genius General Hospital with a week of malaise and progressive shortness of breath. She was found to have diffuse pulmonary infiltrates due to Hantavirus infection. In the ICU she was treated with high-flow nasal cannula, but eventually she tired out and required intubation. Prior to intubation, she was saturating 87% on 100% FiO2 at 60 liters/minute flow. Following intubation, she was placed on APRV. Over a few hours her lungs gradually recruited and she was weaned down to 30% FiO2. Her PaO2/FiO2 ratio on APRV was 475. Later on, due to some concerns regarding hypotension, she was transitioned to conventional low tidal-volume ventilation with 12 cm PEEP. Her oxygenation deteriorated and she required an increase from 30% to 50% FiO2. On conventional ventilation, her PaO2/FiO2 ratio decreased from 475 to 166. She recovered very rapidly and was extubated about 36 hours later. Over the next few days she was weaned off oxygen entirely and walked out of the hospital on room air (1).
Did this woman have ARDS? According to the Berlin Definition shown below, she had moderately severe ARDS while on conventional low tidal-volume ventilation (PaO2/FiO2 ratio of 166). However, she didn't meet the definition of ARDS while she was APRV, a few hours earlier (PaO2/FiO2 ratio 475). Her overall clinical course with prompt recovery and weaning off oxygen is inconsistent with the natural history of ARDS.
Let's perform a thought experiment. Imagine that this woman had been intubated and placed on conventional low tidal-volume ventilation. Without recruitment, it's likely that her PaO2/FiO2 ratio would have quite low. Based on current guidelines, clinicians might decide to paralyze and prone her. Following this management pathway, she wouldn't have been extubated in 36 hours. Her ICU course probably would have been considerably longer– involving deep sedation, paralytics, delirium, and perhaps additional iatrogenic complications.
This case highlights major problems with the ARDS definition, and with our management of “ARDS.” Diagnosis and treatment hinge on the PaO2/FiO2 ratio, which is a function of how the ventilator is set – fluctuating widely over a period of hours.
True ARDS: Increasingly uncommon
“True ARDS” might be defined as a histological diagnosis involving diffuse alveolar damage throughout the lungs (characterized by hyaline membrane formation and thickening of the alveolar walls). These patients clinically behave as follows:
- Severe hypoxemia which doesn't disappear following recruitment.
- Impaired lung compliance.
- Increased dead space (inefficient CO2 clearance, requiring high minute ventilation).
- Slow recovery over a period of several days, often with residual fibrosis.
True ARDS still exists, but it is increasingly rare. Advances in critical care have reduced its incidence:
- Conservative volume resuscitation strategy, with early vasopressors.
- Ongoing monitoring of input/output balance throughout ICU admission, with avoidance of insidious volume overload.
- Use of BiPAP and/or HFNC to avoid intubation.
- Lung-protective ventilation for all
- Targeting light sedation, to avoid obtundation or delirium which increase the duration of intubation.
- Aggressive protocoled extubation to HFNC, to get patients off the vent quickly and keep them off.
- Conservative blood transfusion for non-hemorrhaging patients.
- Hemostatic resuscitation with a 1:1:1 ratio for massive hemorrhage.
In retrospect, most cases of ARDS in the past were probably iatrogenic, due to the following factors:
- Volume overload.
- Volume overload.
- Volume overload.
- Ventilation with inadequate recruitment, causing only a portion of the lung (“baby lung”) to inflate. This delivered the entire stress of mechanical ventilation to a smaller lung volume, causing lung injury. Inadequate recruitment also promotes cyclic opening/closing of alveoli (atelectotrauma).
- Sluggish strategies to liberate patients from the ventilator, such that #4 had sufficient time to cause severe injury.
PseudoARDS: Diagnosis and types
PseudoARDS refers to patients who mimic ARDS, but don't truly have severe lung damage. These patients may technically meet the definition of ARDS. Their outcome is usually surprisingly good, compared to true ARDS patients. Three patterns of pseudoARDS are the most common:
#1: PseudoARDS-Volume pattern
This pattern was initially recognized several years ago (Schrier 1995). The combination of the obesity epidemic and the SEP-1 sepsis core measures in the United States have made this increasingly common recently. The following pattern of events is typical:
- A patient with morbid obesity and moderate pneumonia presents to the hospital.
- 30 cc/kg fluid is bolused to satisfy the SEP-1 core measures for a diagnosis of “sepsis.” For a morbidly obese patient who is receiving other intravenous medications, this causes administration of ~5 liters of fluid over a couple hours.
- The patient deteriorates over the following several hours and requires intubation.
- The patient is supported by the ventilator and diuresed (now that the patient has “ARDS” treatment shifts to a conservative volume strategy). Over 2-3 days, the initial fluid bolus is diuresed off and the patient is extubated.
PseudoARDS-volume patients often have an initial injury causing non-cardiogenic pulmonary edema (e.g. pneumonia, pancreatitis). Subsequently, iatrogenic volume overload is superimposed upon this injury and becomes the major driver of deterioration. The primary lung injury increases the permeability of capillary endothelium in the lung, rendering the lungs more vulnerable to volume overload. Following diuresis, these patients improve much faster than a true ARDS patient would.
#2: PseudoARDS-Effusion pattern
PseudoARDS-effusion refers to a patient with bilateral infiltrates who seems to have ARDS, but whose respiratory failure is actually due largely to effusions and adjacent atelectasis. This may closely mimic ARDS, especially among patients with an admixture of true lung infiltrates plus effusions (e.g. small pneumonia plus large effusions). Large effusions often evade detection in a semi-supine chest X-ray, due to posterior layering of fluid. Diligent use of bedside ultrasonography should detect most of these cases. However, it may be difficult to discern whether the effusion(s) are causing respiratory failure or are merely an incidental finding. CT scan is sometimes helpful to sort this out:
Definitive treatment of pseudoARDS-effusion is drainage of the effusion(s), which causes rapid improvement. Even if effusions are transudative, diuresis may take many days to resolve the effusion. Thus, if the patient is severely distressed or intubated, drainage will accelerate recovery.
#3: PseudoARDS-Collapse pattern
PseudoARDS-collapse is respiratory failure which is largely due to atelectasis. The most common form of this is collapse of the lower lobe(s) or dependent lung tissue, which is a common problem among morbidly obese patients or intubated patients (and especially morbidly obese patients who are intubated).
However, collapse doesn't always follow an easily identified, lobar pattern. For example, the patient discussed above had pseudoARDS-collapse due to diffuse collapse of alveoli throughout her lungs. Her initial CT scan didn't suggest atelectasis. However, she improved dramatically with recruitment, proving that a major cause of her hypoxemia was collapse:
PseudoARDS-collapse is potentially the hardest form of PseudoARDS to diagnose. Unlike PseudoARDS-volume, there is no tell-tale history of heart failure or volume administration. Radiologically, PseudoARDS-collapse can look nearly identical to true ARDS. The only feature that reliably diagnoses PseudoARDS-collapse is resolution with increased mean airway pressure (e.g. APRV or conventional ventilation with high levels of PEEP).
The Berlin Definition attempts to weed out PseudoARDS-collapse by requiring that patients be treated with at least 5 cm PEEP (either invasively or noninvasively). This is a nominal improvement upon the prior definition, which didn't require this (Bernard 1994). However, 5 cm of PEEP is insufficient to cause recruitment (especially among morbidly obese patients). The Berlin committee did consider requiring 10 cm PEEP, but unfortunately rejected it.
PROSEVA definition of ARDS: An improvement over the Berlin definition
The PROSEVA trial was a major RCT which showed benefit from proning. It is summarized here:
Hidden within the fine print of the PROSEVA trial is a new definition of ARDS. Before inclusion in the study, patients were optimized on the ventilator for 12-24 hours. If the PaO2/FiO2 ratio increased to >150 during this period, the patient was excluded from this trial.
Thus, the PROSEVA trial essentially modifies the Berlin definition of ARDS, with the additional requirement that the PaO2/FiO2 ratio remain low despite 12-24 hours of optimization on the ventilator. This improves the specificity of the Berlin definition, because it requires that the PaO2/FiO2 ratio remain stable over time.
It is unclear precisely what the PROSEVA investigators did to optimize patients on ventilation for 12-24 hours. Nonetheless, it's probably a fair assumption that this involved reasonably high levels of PEEP, well above 5 cm. Such measures would sift out patients with PseudoARDS-collapse (who often improve substantially with positive pressure) vs. patients with true ARDS.
The concept of evaluating oxygenation over 24 hours has been proposed before PROSEVA. Villar 2007 studied 170 patients who initially met the definition of moderate ARDS (PaO2/FiO2<200 at PEEP of 9 +/-3 cm)(2). When evaluated 24 hours later at a PEEP>10 cm, only 58% of patients still met criteria for moderate ARDS. Oxygenation often improved dramatically over 24 hours (figure below). This proves that a single measurement of PaO2/FiO2 ratio at relatively low PEEP has terrible specificity for predicting persistent, severe hypoxemia (on par with flipping a coin).
Recently a RCT by Zhou compared APRV vs. low tidal-volume ventilation in ARDS (discussed further on the blog here, and summarized below:
One objection to this study is that the results are too good to be true. This is a single-center study which requires replication, so this is certainly possible. However, another possible explanation for the dramatic benefit from APRV is that a subset of patients in this study had pseudoARDS-collapse. Such patients would be expected to do much better with APRV:
- APRV is extremely effective for recruiting lung among patients with pseudoARDS-collapse. Such patients would experience very rapid improvement in oxygenation, prompting clinicians to consider early extubation.
- Conventional ventilation is less effective at recruitment, especially among patients with collapse of the lower lobes. Thus, patients with pseudoARDS-collapse randomized to conventional ventilation would probably remain on ventilation longer (potentially requiring proning to resolve their atelectasis).
Proposal: Treatment algorithm for ARDS based on PROSEVA & recruitment trial.
For a patient intubated with profound hypoxemic respiratory failure, there are roughly two treatment pathways which seem most common (3):
- APRV, often with light sedation (APRV is more effective at recruiting the lung bases and improving cardiac output when the patient is actively breathing).
- Conventional low tidal-volume ventilation – often with deep sedation, proning, and/or paralysis.
There isn't adequate evidence to definitively determine which pathway is best. Equipoise exists, with different physicians and institutions having varying preferences. For example, many centers in Spain seem keen on proning. Meanwhile, Maryland Shock Trauma in the US seems to favor broader utilization of APRV.
Below is a compromise which might be acceptable to both camps. Following intubation, the patient should be optimized on the ventilator for 12-24 hours (as in the PROSEVA trial). Given the failure of recruitment maneuvers in the ART trial, the safest approach to recruitment may be APRV. APRV allows for gradual recruitment of the lungs over several hours using a high mean airway pressure (~25 cm), rather than a short blast of 40 cm pressure which could cause more hemodynamic instability. It's also notable that some patients with complete lobar collapse may require several hours on APRV to recruit fully – recruitment isn't something that necessarily occurs over 40 seconds in a traditional recruitment maneuver. Following a trial of recruitment, the decision is made whether to pursue proning/paralysis based on PaO2/FiO2 ratio (a strategy which is entirely consistent with best evidence from the PROSEVA and Papazian trials).
Paralysis and proning do appear to reduce mortality in carefully selected patients, but they are fairly involved therapies with definite risks of iatrogenic harm. The above algorithm could promote use of these therapies among patients who truly benefit from them, while simultaneously avoiding therapies in broad swaths of patients who aren't likely to benefit (e.g. patients with PseudoARDS-collapse who improve rapidly with recruitment).
Parting shot: sorry, but many other factors also affect PaO2/FiO2.
Perhaps the most directly studied is FiO2. Increasing the FiO2 tends to increase the PaO2/FiO2 ratio. This occurs because increasing the FiO2 will increase the mixed venous oxygen saturation, which minimizes the impact of shunting deoxygenated blood through the lungs. Villar 2007 showed that increasing the FiO2 to 100% caused ~20% of patients with moderate-severe ARDS to be re-classified into a milder diagnostic category.
Other factors which will affect PaO2/FiO2 ratio include cardiac output, metabolic rate, and hemoglobin concentration. These variables can independently and strongly affect the PaO2 as previously explored here. For example, a drop in cardiac output would decrease the mixed venous oxygen saturation and thereby decrease the PaO2/FiO2 ratio.
Thus, PaO2/FiO2 ratio is a very labile measurement which can be influenced by numerous factors (PEEP, recruitment, FiO2, cardiac output, metabolic rate, and hemoglobin concentration). A more consistent measurement of lung function might be achieved as follows:
- Measurement over time (discussed above).
- Greater effort to account for variations in PEEP and recruitment.
- Measurement of the shunt fraction, rather than the PaO2/FiO2 ratio. Calculation of the shunt fraction using the mixed venous oxygen saturation (or approximating it based on central venous oxygen saturation) could theoretically eliminate much of the variation due to FiO2, cardiac output, hemoglobin, and metabolic rate.
For now, we should bear in mind that PaO2/FiO2 ratio is only one piece of information, which must be interpreted within clinical context. The urge to rest major treatment decisions entirely on a single measurement of PaO2/FiO2 should be resisted. Unfortunately, our definition of ARDS and most clinical research has been built upon this squishy foundation.
- The Berlin Definition of ARDS is limited due to wide fluctuations in PaO2/FiO2 ratio which occur with varying levels of PEEP, different ventilator modes, and different FiO2 levels. This may cause a patient to meet the definition of ARDS one minute, but not the next.
- The incidence of ARDS is decreasing with modern critical care practice. The nature of the disease itself may be changing, since it is largely an iatrogenic phenomenon.
- PseudoARDS refers to ARDS mimics who don't actually have severe lung injury and shouldn't be treated via ARDS pathways. Common forms of PseudoARDS include volume overload, bilateral effusions, and lung collapse.
- The PROSEVA study introduced a more sophisticated definition of ARDS: patients who meet the Berlin definition over 12-24 hours of optimization on mechanical ventilation. In particular, this may help sort out patients with true ARDS vs. patients with atelectasis who respond rapidly to recruitment.
- A diagnostic/therapeutic strategy for ARDS is suggested, based on the PROSEVA trial (figure below). Initial management includes stabilization on the ventilator and determining how well the patient responds to recruitment. Patients with persistent severe hypoxemia have true ARDS and may benefit from proning. Alternatively, patients who improve dramatically with recruitment have pseudoARDS and don't require proning/paralysis.
- Villar J et al. The Berlin definition met our needs: no. Intensive Care Med 2016.
- APRV: Resurrection of the open-lung strategy? (PulmCrit)
- Fighting refractory ARDS with physiologic jujitsu (PulmCrit)
- Diprivan induced pseudo-shock and hypoxic illness syndrome (Scott Aberegg, Medical Evidence Blog)
- Incidentally it's worth noting that this patient was treated with a very fluid-conservative strategy, even despite occasional requirement for vasopressors. She had very exuberant capillary leak, with dramatic hemoconcentration and third-spacing of fluid. My guess is that if she had received a significant amount of crystalloid this would have exacerbated her lung injury substantially. This case raises the question of how much of the morbidity from Hantavirus pulmonary syndrome (quoted fatality rate 50%) is due to iatrogenic harm from crystalloid administration.
- This study was done in 2007 using the prior definitions. I've changed the terminology slightly to be consistent with current definitions. The average PEEP was 9 cm with a standard deviation of 3 cm, so it's likely that nearly all patients had 5cm PEEP or more (thereby meeting the Berlin definition for moderate ARDS).
- There are certainly more possibilities. For example, APRV can be performed in a prone position. In practice, I think these are often the most commonly followed pathways.