Physiologic Markers of Disease Severity in ARDS

Abstract

Despite its significant limitations, the P_aO2/F_IO2 remains the standard tool to classify disease severity in ARDS. Treatment decisions and research enrollment have depended on this parameter for over 50 years. In addition, several variables have been studied over the past few decades, incorporating other physiologic considerations such as ventilation efficiency, lung mechanics, and right-ventricular performance. This review describes the strengths and limitations of all relevant parameters, with the goal of helping us better understand disease severity and possible future treatment targets.

Introduction

ARDS is characterized by lung injury, loss of aerated tissue, increased pulmonary vascular permeability, and inflammatory dysregulation.^1,2 These abnormalities impair gas exchange, lung mechanics, and cardiovascular performance, leading to multiple organ dysfunction. The mortality of patients with ARDS remains high, between 30–50% depending on disease severity.³ Recognition of the disease and implementing best practices such as lung-protective ventilation and prone positioning are suboptimal.³ Clinical risk stratification of ARDS relies on the P_aO2/F_IO2. This parameter only expresses oxygenation failure and may vary based on ventilator settings and administered F_IO2⁴ In addition to oxygenation failure, ARDS results in impairment of ventilation, pulmonary mechanics, and right-ventricular function (Fig. 1); these have also been explored as markers of severity and prognosis. Here we review the evidence and rationale for the studied parameters in each of these domains in the adult population.

Fig. 1.

Physiologic markers of ARDS severity.

Oxygenation

Historically, the classification of disease severity in ARDS has focused on oxygenation failure. Consensus definitions of ARDS rely on the P_aO2/F_IO2.^2,5 This measurement is easily obtainable and has been used in all the recent ARDS trials. However, this calculated variable has significant physiological limitations. The first is the inability to obtain exact measurements of F_IO2 unless the patient is intubated or on a high-flow oxygen device. Second, the P_aO2/F_IO2 has a curvilinear relationship with F_IO2 and will greatly vary according to the F_IO2 administered without a change in disease severity. That is, the presence of shunting and ventilation/perfusion mismatch affects the P_aO2 variation in response to F_IO2 changes, leading to different P_aO2/F_IO2 for a given patient. One study showed disease severity classification changing for about 30% of subjects when testing different F_IO2 levels across typical ranges of arterial oxygen saturation (92–98%).⁴ PEEP also modifies P_aO2, leading to higher P_aO2/F_IO2 for patients on higher levels of PEEP compared to lower levels of PEEP.^6-8 The effect of PEEP in recruiting collapsed alveoli and improving oxygenation can rapidly modify P_aO2/F_IO2 even on day 1 of illness, sometimes enough for patients to be reclassified to a lower severity or not meet criteria for ARDS at all. This phenomenon of rapidly improving ARDS is widely prevalent,⁹ and trials have tried to address this issue by only including subjects who still meet the specified P_aO2/F_IO2 criteria after 12–24 h of mechanical ventilation.^10,11 Other interventions that increase P_aO2, such as pulmonary vasodilators, will also affect the P_aO2/F_IO2 calculation.^12-16 Finally, several studies have demonstrated that initial P_aO2/F_IO2 values are not independently associated with mortality,^17-23 although measurements obtained on subsequent days may correlate better with outcomes.^24,25

Considering these limitations, other oxygenation markers have been studied, including the oxygenation index (OI),^18,26-28 the less invasive oxygenation saturation index (OSI),^29,30 and the P_aO2/F_IO2 PEEP (P/FP) ratio.³¹

P/FP Ratio

The Berlin definition requires a minimum PEEP of 5 cm H₂O to diagnose ARDS. However, higher levels of PEEP are often used in clinical practice. Therefore, the P/FP ratio incorporates set PEEP as follows:³¹

$$P/FP= \frac{P_{a{O}_2}\times 10}{F_{I{O}_2}\times PEEP}$$

A retrospective study of 3,442 subjects from 7 randomized trials in ARDS demonstrated that P/FP reclassified 1,860 subjects (54.0%) to a different disease severity than P_aO2/F_IO2.³¹ When compared with P_aO2/F_IO2, the P/FP was better at predicting mortality (Table 1), with the prognostic performance improving with each increasing level of PEEP tested. Other than this, supportive data on the P/FP ratio are limited, and it is unclear how it performs in patients with altered chest-wall compliance (such as severe obesity).³²

Table 1.

Studies Evaluating Severity of Oxygenation Derangements on Outcomes in ARDS

Oxygenation Index

OI was initially used to assess the severity of hypoxemic respiratory failure in neonates and the need for extracorporeal membrane oxygenation (ECMO).³³ It is calculated as follows:

$$OI= \frac{F_{I{O}_2}\times mean P_{aw}\times 100}{P_{a{O}_2}}$$

where P_aw represents airway pressure. It has prognostic value in adult patients with ARDS, with studies showing OI to have a moderate and consistent association with mortality,²⁶ predictive value on day 3 for the outcome of death or failure to wean,²⁷ and a statistically significant association with mortality after multivariable analysis.¹⁸

An ARDS cohort was analyzed based on OI, P_aO2/F_IO2, and the different severity levels of the American-European Consensus Conference definition and the Berlin definition, with all of them showing association with mortality on univariable analysis but not on multivariate analysis.²⁸ The studied criteria were progressively more accurate from day 1–7. Day 3 values were chosen as an ideal combination between earlier prognostication and better accuracy compared to day 1. In a stepwise backward selection model, OI was the only variable retaining statistically significant value for predicting mortality.

A modification of the OI formula, the adult age-adjusted OI (AOI), is calculated by adding age to OI. The AOI seems to perform better at predicting mortality²⁶ (Table 1). However, AOI has not been evaluated in further studies, and more evidence would be needed to confirm its prognostic value.

Overall, the association of OI with mortality varied across these studies, and discrimination values were modest (area under the curve 0.610–0.724).^26,27 It may perform better than P_aO2/F_IO2 alone due to the inclusion of mean airway pressure, a manifestation of the respiratory system characteristics, and ventilator settings.

Oxygenation Saturation Index

OSI is a variation of the OI formula incorporating noninvasive S_pO2. It is calculated as:

$$OSI = \frac{F_{I{O}_2}\times mean P_{aw} \times 100}{S_{p{O}_2}}$$

S_pO2 must be ≤ 97% as higher measurements have lower discrimination values and were excluded from studies.^29,30

A single-center retrospective cohort of 329 subjects with ARDS analyzed OSI, OI, P_aO2/F_IO2, S_pO2/F_IO2, and lung injury score on day 1 and showed that OI and OSI have a strong correlation (Spearman rho 0.862, P < .001).²⁹ Only OSI was found to be independently associated with mortality after multivariable analysis, while no independent association was found for the other criteria. Another retrospective study of 101 subjects with ARDS found both OI and OSI to have predictive value, with a threshold of OSI ≥ 12 and OI ≥ 16 associated with increased mortality.³⁰ OSI has the benefit of not requiring a blood gas sample; however, there are fewer studies to support it; its discrimination was modest, with an area under the curve ranging from 0.602–0.656;^29,30 and pulse oximetry has been shown to be less accurate in Black patients, potentially leading to racial bias.³⁴

Ventilation and Dead Space

ARDS also leads to decreased ventilation efficiency due to increased dead space (V_D) relative to the tidal volume (V_T). Dead space in ARDS is attributed to regional hyperinflation, alveolar and endothelial damage, microvascular thrombi, and disordered pulmonary blood flow.^35,36 In addition, the disease severity, hemodynamics, mechanical ventilation type, and its intensity modify dead space.

Dead-Space-to-Tidal-Volume Ratio

The ratio of dead space to tidal volume (V_D/V_T) requires measurement of the expired CO₂ and arterial CO₂ and using the Enghoff modification of the Bohr equation:³⁷

$$V_D/V_T = \frac{P_{aC{O}_2}– P_{\overline{E}C{O}_2}}{P_{aC{O}_2}}$$

where P_E̅CO2 is defined as the partial pressure of CO₂ in mixed expired gas.

Several studies have demonstrated an association of V_D/V_T with mortality^37-43 (Table 2). A mean V_D/V_T > 0.58–0.60 was consistently seen in subjects who died.^38,40 The calculation of the V_D/V_T requires the P_E̅CO2 for which volumetric capnography is needed and not routinely available. Therefore, a bedside estimation of V_D/V_T was proposed using the end-tidal CO₂ (P_ETCO2) in the Enghoff-Bohr equation:⁴¹

$$V_D/V_T=\frac{P_{aC{O}_2}-P_{ETC{O}_2}}{P_{aC{O}_2}}$$

Table 2.

Studies Evaluating Ventilatory Parameters and Outcomes in ARDS

This simplified method was associated with mortality on multivariate analysis on a retrospective study of subjects from the ARDS Network database.⁴¹ In this study, the measurements were obtained within 24 h of the diagnosis of ARDS, and P_ETCO2 was recorded within 1 h of the arterial blood gas with no modifications in the ventilator settings.

Ventilatory Ratio

Ventilatory ratio (VR) is a simpler method to estimate dead space, which only requires P_aCO2 and the measured minute ventilation:^42-45

$$VR= \frac{minute ventilation (mL/min)\times P_{aC{O}_2} }{predicted body weight (kg) \times 100\times 37.5}$$

VR was studied in 520 subjects with ARDS and was found to have a moderate positive correlation with V_D/V_T (modified r = 0.66).⁴⁴ A VR < 2 was associated with lower mortality in the primary data set and in the validation cohort extracted from the FACTT trial.⁴⁶ A similar cutoff for VR was demonstrated in 1,307 subjects, where a VR value > 2.06 was associated with mortality.⁴² An analysis of 940 subjects with ARDS found both VR and estimated V_D/V_T to be independently associated with mortality on multivariable analysis.⁴³ Univariable analysis of 340 subjects from a single institution showed VR to have a weak but statistically significant predictive value for overall mortality.⁴⁵

V_D/V_T and VR seem to both have prognostic value in patients with ARDS, with recent studies still exploring the clinical value of VR. In patients with COVID-19 receiving mechanical ventilation in the ICU, a higher VR and its increase at day 3 were associated with higher mortality.⁴⁷ When analyzed to guide extracorporeal CO₂ removal, an elevated VR (≥ 3) was associated with a survival benefit from this intervention, while a low VR (< 3) was associated with increased likelihood of harm.⁴⁸

Lung Compliance, Driving Pressure, and Mechanical Power

Another focus area is physiological markers of the respiratory system characteristics (eg, compliance) and the forces exerted on the respiratory system by the ventilator (eg, driving pressure [ΔP] and power) as a manifestation of the severity of disease or a potential cause of lung injury (Table 3).⁴⁹

Table 3.

Studies Evaluating Lung Mechanics and Outcomes in ARDS

Lung Compliance

Respiratory system compliance (C_RS) is often decreased in ARDS.⁵ Therefore, it has been studied as a possible prognostic marker shown to be slightly lower in non-survivors.^18,28,37,43 However, its use as a mortality prognostic tool had a poor predictive value,^45,50 and an attempt to improve C_RS with recruitment maneuvers and PEEP titration was not shown to be beneficial.⁵¹

There are several challenges with the use of C_RS, for one, its dependance on lung volume, chest-wall mechanics, and size of the lung. Its use alone has not panned out as a good prognostic or severity marker, albeit very low values are related with increased mortality.

Concepts such as ΔP⁵² and mechanical power (MP)⁵³ normalize the forces exerted by the V_T delivered by the ventilator to the lung tissue available for ventilation. Given the variability in size and heterogeneity of the aerated lung in ARDS, ΔP and MP can better represent the process of lung damage from volutrauma, characterized by excessive alveoli distention (strain), and barotrauma, characterized by excessive transpulmonary pressure (stress).

Driving Pressure

ΔP is defined as the ratio of V_T to C_RS, calculated as plateau pressure (P_plat) minus total PEEP. It was studied by Amato et al⁵² as an important mediator of mortality in ARDS. Since its concept normalizes V_T to C_RS, it is proposed to better represent the dynamic forces responsible for ventilator-induced lung injury in patients with ARDS. The study analyzed ventilatory variables from 3,562 subjects from previous randomized clinical trials and found a strong and independent association of ΔP with mortality.⁵² A subsequent study analyzed ΔP in 787 subjects from two ARDS trials that used strictly lung-protective ventilation as standard of care on both arms.⁵⁴ They found a hazard ratio (HR) of 1.05 (1.02–1.08) for mortality at day 90 per each cm H₂O increase in ΔP on multivariate analysis. This was similar to the findings from the Amato et al study,⁵² which found a 41% increase in mortality for a 1 SD (about 7 cm H₂O) increase in ΔP, corresponding to an HR of mortality of 1.049 per each cm H₂O increase in ΔP.

Goligher et al⁵⁵ analyzed data from subjects included in 5 randomized trials and showed that the mortality reduction from low V_T ventilation strategy depends on the normalized respiratory system elastance (normalized to ideal body weight). This analysis showed that lower V_T was only associated with decreased mortality in subjects with higher elastance (those who would inherently have a higher ΔP with conventional V_T values), while those with low elastance did not seem to benefit from the low V_T strategy. A reanalysis of the ALVEOLI⁵⁶ and ExPress trials⁵⁷ investigated the changes in ΔP and P_aO2/F_IO2 (ΔP_aO2/F_IO2).⁵⁸ When combined in the same model, change in ΔP after randomization had a stronger association with mortality than ΔP_aO2/F_IO2, suggesting that the concept of ΔP could be useful not only as a prognostic tool on day 1 but also as a possible target for ventilator adjustments. Of note, several studies have not found ΔP predictive of mortality in subjects with obesity.^59-61

Mechanical Power

MP is a measure of the energy delivered to the respiratory system by the ventilator.⁵³ It incorporates flow, breathing frequency, PEEP, V_T, and ΔP. An equation for calculating MP in J/min during volume control ventilation with constant inspiratory flow was defined as:

$$M{P}_{VCV}= 0.098\times RR \times \left\{V_T{}^2\times [ \frac{1}{2} \times E{L}_{RS}+RR \times \left(\frac{1+I:E}{60 \times I:E} \right)\times R_{aw} ]+V_T\times PEEP\right\}$$

where I:E is the inspiratory-to-expiratory time ratio, and R_aw is the airway resistance.⁵³ It has been shown to correlate extremely well with the actual measurement of MP in subjects with ARDS in the research setting (R² 0.96 for PEEP 5 cm H₂O). The calculations assume a linear relationship between lung volume, elastance, and resistance, perhaps less accurate at the extremes of respiratory mechanics. There are several equations for different ventilator modes and waveforms.

Equations for pressure control ventilation are similarly accurate compared to a reference measurement of MP.⁶² These equations are generally too complicated to be used at bedside. Given the complexities of the available formulas, a simplified equation was proposed:

$$MP= \frac{{\dot{V}}_E \left(Peak pressure+PEEP+F/6\right)}{20}$$

where V̇_E represents minute ventilation in L/min and F represents constant inspiratory flow in L/min; it was shown to have a very strong correlation to the original equation, with R² ranging from 0.97–0.99.⁶³ In two other cohorts, MP was independently associated with in-hospital mortality. Risk of death increased consistently above 17 J/min even in subjects ventilated with low V_T.⁶⁴ In another retrospective study with 222 subjects with ARDS, MP was not different between survivors and non-survivors (14.97 [11.51–18.44] J/min vs 15.46 [12.33–21.45] J/min) and did not affect ICU mortality; when normalized to C_RS and well-inflated tissue (as seen on chest computed tomography), MP was associated with increased ICU mortality.⁶⁵

A comparison of MP with primary ventilator variables such as breathing frequency, V_T, ΔP, PEEP, and P_plat in 4,549 subjects from 6 clinical trials and one large observational cohort showed an association with death after multivariable adjustment that was only significant for the elastic dynamic component of MP, which is related to ΔP, and not for its elastic static component, related to PEEP, or for its resistive component; among ventilator variables, only ΔP and breathing frequency were independently associated with death.⁶⁶ There was no significant interaction between ΔP and breathing frequency (P = .08), and the effect of each 1 cm H₂O increase in ΔP was approximately equivalent to 4 times that of each 1 breath/min increase in breathing frequency. In the same study, a combined model with ([4 x ΔP] + breathing frequency) was associated with mortality and maintained its predictive value even in a multivariable model with MP.

There is no prospective randomized evidence that a ventilatory strategy targeting ΔP or MP is superior to the ARDS Network protocol.⁶⁷ However, it is a simple bedside calculation that predicts mortality in ARDS⁵² and may add value in the future by allowing individualized ventilator prescriptions for patients based on lung mechanics.⁵⁵

Right-Ventricular Injury

Finally, the effect of ARDS on the cardiovascular system, through increased pulmonary vascular resistance, vasoconstriction from hypoxemia, microthrombi formation, vascular remodeling, extrinsic vessel occlusion by high PEEP/P_plat/edema, and the eventual vascular remodeling in prolonged courses,⁶⁸ may lead to right ventricle (RV) injury. A meta-analysis and systematic review of 9 studies including a total of 1,861 subjects showed a 21% prevalence of RV injury in subjects with ARDS, increasing odds of mortality by 45% (odds ratio 1.45 [95% CI 1.13–1.86]).⁶⁹ The definition of RV injury varies significantly across studies as outlined in Table 4. Because the risk factors for RV injury in ARDS include variables that also suggest more severe lung injury (such as hypercapnia and elevated ΔP),⁷⁰ it is unclear whether the presence of RV injury is contributing directly to mortality or simply serving as an indicator of disease severity.

Table 4.

Studies Evaluating Right-Ventricular Dysfunction and Outcomes in ARDS

Discussion

Prognostication of patients with ARDS remains challenging. P_aO2/F_IO2 is still the main parameter used in clinical care and research despite its known limitations, especially on day 1.^17-23 Reviewing the different prognostic criteria that were developed through the last decades gives us the opportunity to better understand the clinical manifestations of ARDS and their relationship to the disease process. The methods analyzed in this review all have predictive value even on day 1, with differing levels of discrimination and ease of application. Many of these variables have improved prognostic value after day 1 of disease, including P_aO2/F_IO2.^24,25

The initial focus of prognostication was on oxygenation and ventilation abnormalities, which are immediate consequences of the primarily pulmonary disorder that is ARDS. They do not seem though to be direct mediators of mortality. Subjects in the low V_T arm of the ARMA trial had better survival despite lower P_aO2/F_IO2 values,⁶⁷ and interventions designed to increase oxygenation such as inhaled vasodilators have failed to show improvement in mortality.^12-16 Furthermore, attempts to optimize oxygenation can impair cardiac output and oxygen delivery, which is unaccounted for in all variables considered in this review. V_D/V_T and VR reasonably capture the ventilation deficits present in ARDS. Their prognostic performance, however, seems to be limited, likely reflecting their indirect relation with the disease process. It is also uncertain whether these values are useful for guiding treatment given the multiple potential causes of dead space in ARDS, as well as a lack of prospective data (Table 5).

Table 5.

Cutoff Values for Prognostic Criteria Used in ARDS Studies

Variables such as ΔP and MP relay useful prognostic information,^49,71 and their use may promote a better understanding of the interaction between ventilator prescription and lung mechanics. ARDS is not a disorder restricted to the lungs, and attention is necessary to all the systems affected by it, especially the cardiovascular system. Right-ventricular injury is prevalent in ARDS and is associated with increased mortality.⁶⁹ With the growth of echocardiography in critical care medicine, RV injury can be more readily recognized. Limitations include the several different criteria used to define RV injury, the reliance on adequate echocardiographic images, and the lack of a clear severity scale and therapeutic targets.⁷²

As a final observation, this review has not focused on the specifics around the use of these prognostic markers in SARS-COV2–related ARDS. We feel that given the extensive literature discussing the utilization of prognostic markers in COVID-related ARDS it is beyond the scope of the present review. This review also did not capture what variables might be useful to prognosticate outcomes in ARDS patients already receiving extracorporeal life support (ECMO). Most scoring systems for this application consider pre-cannulation variables, which may be less relevant after cannulation since gas exchange is directly altered on ECMO, and ventilator settings are prescribed differently. Consequently, these prognostic markers might not be very helpful in specific populations such as patients undergoing ECMO or in morbidly obese patients. Future studies need to specifically address the knowledge gap related to these specific populations.

Summary

A description of the physiology and severity of ARDS cannot be captured in any single parameter. The cause of death in patients with ARDS is often related to extra-pulmonary organ failures,⁷³ making it difficult for respiratory variables to account for the totality of morbidity and mortality risk. Prospective studies are also lacking. Nevertheless, many of the parameters described in this review outperform P_aO2/F_IO2 in terms of classifying disease severity, offering opportunities to refine ARDS definitions for future research.