Problems with ARDS definitions

Clinical vignette

A patient is hospitalized with worsening sepsis secondary to a urinary tract infection and develops dyspnea, hypoxemia and increased respiratory effort with radiographic evidence demonstrating new diffuse pulmonary infiltrates. The patient is transferred to the intensive care unit (ICU) where clinicians commenced on high-flow nasal oxygen (HFNO). After several hours, the work of breathing remains elevated and there is SpO2 90% despite a HFNO at 50 L/min. As such, the patient is intubated and connected to mechanical ventilation (MV) with a tidal volume (VT) of 7 ml/kg predicted body weight (PBW) and a positive end-expiratory pressure (PEEP) of 12 cmH2O. The patient’s PaO2 increases to 160 mmHg with a FiO2 0.5 (PaO2/FiO2 ratio 320 mmHg). Rapid improvement was noted following administration of antibiotics, fluids and light sedation. The patient was successfully extubated after fifty hours of MV and discharged from hospital a few days later.

Case discussion

Did this patient have acute respiratory distress syndrome (ARDS)? According to the current Berlin definition [1], this patient met the criteria for moderate/severe ARDS, based on the acuity of presenting symptoms, the radiographic evidence of bilateral pulmonary infiltrates and the initial SpO2/FiO2 ratio when receiving HFNO therapy. The patient, however, no longer met diagnostic gas-exchange criteria after only a few hours of MV. Such a rapid recovery is conceptually inconsistent with the natural history of ARDS. This case serves to highlight several major issues with the current ARDS definition and its management. Firstly, the PaO2/FiO2 ratio is largely a function of ventilator settings [2]. Secondly, it is plausible that the PaO2/FiO2 ratio on MV would have been below 150 mmHg had the clinicians opted for a PEEP < 12 cmH2O. A ratio of this level may have prompted the clinicians to escalate the respiratory support for use of neuromuscular blocking agents to paralyze the patient or use of prone positioning. The apparent ‘need’ to utilize these techniques would likely delay the patient’s weaning and extubation while increasing their risk of iatrogenic complications. A single measurement of PaO2/FiO2 on admission, prior to any treatment optimization particularly if at relatively low PEEP, as indicated by the Berlin definition [1], has shown poor performance for predicting ARDS severity [3] (Table 1).

Table 1 Limitations of the current definition and diagnostic/therapeutic approach of ARDS
Full size table

Background

The condition subsequently identified as ARDS has been known for over a century, but the first summary description of this heterogeneous pulmonary disorder was published in 1967 [4]. The clinical features included severe dyspnea, hypoxemia, decreased lung compliance and diffuse alveolar infiltrates on the chest X-ray, in a setting where cardiogenic pulmonary edema had been ruled out. Since this first description, the ARDS definition has been revised several times while many researchers and clinicians questioned its existence as a discrete entity [1, 5,6,7]. Authors of each revision [1, 6, 7] justified their selected criteria by pointing out the flaws in the previous definition and pledged that the “new” definition would be able to solve past shortcomings.

Each definition used the PaO2/FiO2 ratio as the main defining criterion for establishing the diagnosis and severity of the syndrome. While PaO2 is the most direct measurement of oxygenation status in ARDS, it is expressed in terms of PaO2/FiO2 ratio both in the AECC and Berlin definitions [1, 7]. There are no data linking PaO2 on a set FiO2 with a wide variety of ventilation settings and modes, to predictable structural changes in the alveolar-capillary membrane or to the extent of diffuse alveolar damage (DAD) at the time of ARDS diagnosis [8]. On the contrary, there is recent evidence showing a correlation between the severity of lung injury and outcome when the PaO2 is measured under standardized ventilatory settings [3]. Other factors affecting PaO2/FiO2 ratio include cardiac output, intrapulmonary shunt fraction, metabolic rate and hemoglobin concentration [9]. Therefore, if PaO2/FiO2 ratio is crucial to ARDS definition and its management, it should be argued that resting clinical decisions on a single value obtained outside a defined standard setting should be rejected [10]. A fundamental problem with the definitions based on criteria with such significant limitations is that operationalizing their application may affect the therapy that patients receive, or if they are enrolled into clinical trials [11], particularly in many hypoxemic patients who improve after 24 h of standard intensive care [3, 10].

The pseudo-ARDS scenario

Various types of pulmonary and systemic insults can lead to a common pathophysiological response [12]. Regardless of the precise mechanism, the typical anatomopathological feature of ARDS is DAD [13]. In general, it is useful to think of the pathogenesis as the result of two different pathways: a direct insult to alveolar cells and an indirect insult to the endothelial cells by an acute systemic inflammatory response. The early exudative phase of DAD is characterized by inflammation and protein-rich edema [13], atelectasis and structural damage to the lung architecture if inflammation persists. Eventually, these changes evolve into a fibroproliferative phase with capillary thrombosis, lung fibrosis and neovascularization. Most ARDS patients die during this phase despite ventilatory and extracorporeal organ support.

Although there are no typical ARDS patients, it is likely that DAD is present in all of them, despite reports showing absence of DAD in a marked proportion of autopsies in patients fulfilling the Berlin criteria for ARDS [14]. This is a likely result of incorrect classification, as in those reports, lung biopsies were performed days or weeks after ARDS onset and/or initiation of therapy, and a lack of randomization in pathological studies makes difficult to determine the correlation between clinical and pathological findings. In addition, lung tissue samples reporting clinicopathological comparison with DAD [15], were obtained from patients ventilated with injurious MV settings with VT up to 16 ml/kg actual body weight [16] or PEEP from 0 to 5 cmH2O in most patients [17]. Criteria that are necessary for a definitive diagnosis of ARDS (increased pulmonary vascular permeability and DAD) are difficult to incorporate into clinical practice. Probably, a simple measure of vascular permeability at the bedside, such as extravascular lung water, is needed in future ARDS definitions for identifying ARDS, although how abnormal must pulmonary vascular permeability be before predicting the presence of DAD is not clearly known [8].

Many forms of acute hypoxemic respiratory failure mimic ARDS and do not have DAD, if one considers how prevalent are fluid overload, bilateral pleural effusions and bilateral atelectasis in ICU [18]. Patients with these features may meet the Berlin definition, but their overall outcome is usually better compared to true ARDS. Enrollment of patients with rapidly improving ARDS or pseudo-ARDS may contribute to the failure of therapeutic clinical trials [19], paving the way to studies where physiological enrichment is used to overcome this issue [2]. Severe hypoxemia caused by lobar consolidation is frequently treated as ARDS, when it is possible that specific treatment options would benefit these patients, while they could be spared from the development of ventilator-induced lung injury (VILI) in the unaffected lung [20].

Problems with hypoxemia

An integral part of the supportive therapy for ARDS is the application of respiratory support aimed at achieving adequate gas-exchange and tissue oxygenation without further damaging the lungs [20]. The use of MV is vital for most ARDS patients, but over the last decade, ARDS patients with mild or moderate forms of lung injury have successfully been managed without endotracheal intubation [11], as recognized by the Berlin definition [1] and by recent guidelines [11].

We suspect that PaO2/FiO2 ratio will be not eliminated from future definitions of ARDS. Of note, a standardized level of FiO2 and PEEP has never been a condition for defining hypoxemia under MV. In patients fulfilling ARDS criteria, assessment at 24 h on PEEP ≥ 10 cmH2O with FiO2 ≥ 0.5 for 30 min caused PaO2/FiO2 ratio to increase, such that more than a third of patients no longer met ARDS criteria [3]. In addition, the exact FiO2 is difficult, if not impossible, to be determined in patients on non-invasive ventilation or HFNO. We suspect that none of proposed indices of oxygenation for ARDS categorization and prediction of outcome will be useful to make clinical decisions unless assessed or calculated using standardized ventilatory settings [21, 22]. In the latest iteration of the definition, some authors have proposed the use of SpO2/FiO2 ratio, mainly keeping in mind the resource constrained environments, where arterial blood gas analysis might be difficult or impossible to achieve [11]. Unfortunately, SpO2 is affected by several variables [23] such as changes in temperature, pH, PaCO2, concentration of 2,3-diphosphoglycerate and carboxyhemoglobin, and its measurement is influenced by ethnicity [24], although none of these variables affect PaO2. SpO2/FiO2 ratio contains all the problems of the PaO2/FiO2 ratio, with the added problem that the 95% confidence interval for SpO2 vs. SaO2 is ± 5% when patient is desaturated, and PaO2 values could fluctuate > 300 mmHg when SpO2 is ≥ 97%.

In the European Collaborative Study [25], the mortality of patients with PaO2/FiO2 < 150 mmHg at 24 h was almost double the mortality of patients with PaO2/FiO2 ≥ 150 mmHg. Three recent clinical trials used a value of PaO2/FiO2 < 150 mmHg at PEEP ≥ 5 [26, 27] or ≥ 8 cmH2O [28] to enroll patients during the first 24–48 h of ARDS diagnosis. It is plausible that in a substantial proportion of patients in recent clinical trials, the severity of lung injury was modest. If patients have a low risk of the condition to be prevented, any trial will not validate the value of the intervention under study [29]. In a recent study with 1303 moderate/severe ARDS patients [2], almost half of them had a PaO2/FiO2 ≥ 150 mmHg at 24 h and their ICU mortality was about 20%, whereas patients with PaO2/FiO2 < 150 mmHg had an ICU mortality greater than 45%. It is possible that in the new updated ARDS categorization, a new PaO2/FiO2 threshold could be incorporated (Table 2).

Table 2 Potential recommendations for improving the definition of ARDS
Full size table

Future directions

We believe that the term ARDS should be used with greater care. As suggested by experts in the field of critical illness, we believe that the current ARDS-based framework of illness should be reconsidered [30]. Clinicians should be interested in operational definition criteria that can trigger the use of therapies with high probability of resulting in improved outcomes (Table 2). To quantify accurately the severity of ARDS, we would ideally need two indices of severity: one that measures the severity of lung injury per se, and another that measures the overall severity of patient’s overall illness which would then quantify the context within which ARDS develops [8, 31]. Without those measures and understanding the effect of specific etiologies on the outcome (Fig. 1), any new updated definition of ARDS will be a perpetual iteration of the same shortcoming without a substantial advancement since its first description [32]. Subdividing ARDS patients into categories reflecting different severities or modifiable pathophysiological processes represents the most critical advance for precision medicine in ARDS. It provides a rationale for identifying patients that are resistant to therapy, or who should be the target for aggressive and innovative therapies, or in whom endotracheal intubation and MV could be avoided, or who should be excluded from some clinical trials [33,34,35]. Most studies on sub-phenotypes in ARDS to date are based on retrospective analyses [36] and it is unclear whether those subtypes of patients represent categorization of the etiologic underlying disease or of ARDS itself [30, 37]. Even with this caveat, it is possible to combine information obtained from lung imaging and pulmonary/systemic biomarkers to personalize individual management of ARDS [38].

Fig. 1
figure 1

The acute respiratory distress syndrome (ARDS) high-speed train showing variables and factors affecting definition and outcome of patients with ARDS. Abbreviations: ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; LPV, lung protective ventilation; MV, mechanical ventilation; PEEP, positive end-expiratory pressure; P/F, PaO2/FiO2 ratio; RCT, randomized controlled trials; SOFA, sequential organ function assessment; SVS, standardized ventilator settings; VD/VT, alveolar dead space; VT, tidal volume

Full size image

ARDS is frequently associated with hemodynamic instability, one of the main determinants of mortality. There is a place for invasive hemodynamic monitoring in patients who need an accurate assessment of their cardiovascular status, although the specific monitoring should be individualized. Vascular alterations in ARDS include vasoconstriction and vasodilation of pulmonary vessels leading to unfavorable blood flow distribution, pulmonary hypertension and right ventricular dysfunction [39, 40]. Management of intravenous fluids and vasopressors in ARDS is a key challenge and a top research priority. One should consider the risks and benefits in each phase of ARDS and facilitate fluid removal. As reported in a recent study, clinicians administer higher doses of fluids and lower doses of vasopressors than recommended by a machine learning (ML) model [41]. Of note, patients receiving doses similar to those recommended by the ML model had the lowest mortality rate.

Greater emphasis should be placed on the role of carbon dioxide (CO2) and dead space (VD/VT) in determining the severity of disease [42]. VD/VT or wasted ventilation (the portion of VT that does not participate in gas-exchange) is not included in any definition of ARDS (Table 1). Elevated VD/VT is associated with lower probability of being discharged alive [43, 44]. The lack of precise information about the anatomic state of the pulmonary vascular circulation makes difficult to establish a rational criterion for ARDS stratification and for initiating specific therapy. Analysis of expired CO2 kinetics provides important non-invasive cardiorespiratory information for clinical assessment, monitoring and management of ventilated ARDS patients. The concept of VD/VT is clinically useful not only to assess and adjust alveolar ventilation during MV but also to detect alveolar overdistension [42].

We do not know yet whether favoring early spontaneous ventilation in ARDS improves outcome when compared to controlled MV plus sedation and proning [45, 46]. In managing ARDS, the underlying disorders lead to a high respiratory drive and should be addressed immediately following intubation. Allowing early spontaneous breathing as soon as some improvements occur could decrease duration of MV. Early spontaneous breathing could allow to use high levels of PEEP to prevent atelectrauma and inflammation for enhancing the lung to heal [46].

Finally, future research should address precision medicine in ARDS, invoking the concept of treatable traits [30]. We need clinical trials comparing current management with that derived from precision medicine. No tools currently exist to personalize treatment of ARDS and assist clinicians in making decisions in real time at the bedside. Features of a ML model to predict ICU mortality suggested that they were clinically interpretable and relied primarily on sensible clinical and biological parameters [31].