BIOMARKERS IN THE ACUTE RESPIRATORY DISTRESS SYNDROME: PAST, PRESENT, AND FUTURE

Abstract

Acute respiratory distress syndrome (ARDS), originally described in 1967, affects more than 3 million individuals each year throughout the world and accounts for approximately 10% of all admissions to the intensive care unit. Despite substantial progress in defining the epidemiology and pathogenesis of the syndrome, there is no specific treatment and mortality rates remain high. Barriers to finding specific therapeutic interventions include the inability to predict who will get ARDS, inadequate definitions and specific diagnostic markers, the heterogeneity of the patient population, complexities of the pathogenesis, and the impact of clinical care. Measurements of biomarkers have identified these barriers as well as contributed to the current understanding of the disease. The COVID-19 pandemic resulted in a dramatic increase in patients with ARDS, driving an urgent need to understand the pathogenesis and develop and implement therapeutic interventions. Past studies of biomarkers in ARDS can provide insight that could help to meet those needs more rapidly.

INTRODUCTION

Adult respiratory distress syndrome (ARDS) is a global problem without a cure. Worldwide, prior to the pandemic, there were almost 3 million patients accounting for 10% of ICU admissions and 24% of the patients in the ICU receiving mechanical ventilation annually. In the United States alone, about 200,000 patients accounted for approximately 75,000 deaths each year (1). Since the onset of the COVID-19 pandemic, the incidence has increased dramatically with the estimated number of cases at almost 25 million (2). This rapid increase in cases drove an urgent need to both understand the pathogenesis of the syndrome and to develop and implement therapeutic interventions. It also provided an opportunity to review the progress to date in ARDS, learn lessons from the past to avoid previous pitfalls, and move ahead more smoothly. A review of the role of biomarkers in ARDS highlights this opportunity.

ARDS was first described in 1967 in a report of 12 patients with similar clinical features including severe dyspnea, tachypnea, refractory cyanosis, decreased pulmonary compliance, diffuse alveolar infiltrates on chest radiographs, and evidence of atelectasis, vascular congestion, hemorrhage, pulmonary edema, and hyaline membranes at autopsy (3). As more cases were recognized, it became clear that the clinical diagnosis of ARDS was complex. Furthermore, lung biopsies were not clinically indicated in most patients, so a search began for biomarkers. Although the initial focus was on a diagnostic biomarker, it quickly became evident that biomarker measurements could also potentially be useful to identify patients at risk for ARDS and predict the development of the syndrome; identify patients with heterogeneous clinical phenotypes; serve as markers for disease severity, prognosis, and outcomes; serve as surrogate end points for trials; influence therapeutic decisions; and elucidate or confirm pathogenesis. The evolution of the role of biomarkers started shortly after the syndrome was first described.

WHY DOES EVERYONE GET DIFFERENT RESULTS?

Early biomarker studies were based on animal models and generally limited to small studies of a single biomarker, measured from a single site (i.e., plasma or bronchoalveolar lavage), in a single clinical center. Frequently, these studies produced conflicting and confusing results. Subsequent investigators identified a number of confounders contributing to the differences in results. These included the lack of concordance between clinical definitions of ARDS, the challenges of the clinical definitions of the identified risk factors such as sepsis, the heterogeneity of the patient population, the selection of appropriate control groups, timing of the studies, the assay methods, and, ultimately, differences in clinical care.

The Definition

The definition of ARDS has been an ongoing quandary. Three early definitions were commonly used: lung injury score, modified lung injury score, and American-European Consensus Conference. Our group compared these three definitions and found there was indeed a lack of concordance among them highlighting the need for a better or more specific definition (4). In 2012, the current definition, the Berlin definition, was published (5). It includes timing, chest imaging, the origin of edema, and oxygenation. The oxygenation level separates patients into three groups: mild, moderate, and severe. Although this definition has been widely used in practice, clinicians still miss the diagnosis frequently. Studies have consistently shown that the syndrome is clinically recognized only about 60%–65% of the time (6). Some of this can be attributed to challenges with the interpretation of chest radiographs (1). The definition itself also likely leads to under recognition of the syndrome. One of the requirements for the diagnosis of ARDS is the use of positive pressure ventilation, thus excluding those patients on noninvasive ventilation and, an increasingly larger group, those patients on high flow nasal cannula.

Patient Heterogeneity

Many distinct clinical risk factors for the development of ARDS have been identified (7), but the definitions of those are also complex. For example, the definition of sepsis, a major risk factor, continues to evolve and, in itself, is a heterogeneous syndrome. Trauma, another major risk factor, is also complicated. Whether or not head trauma and long bone fractures are categorized together as a homogeneous group under the heading “trauma,” whether long bone fractures are included in or separate from the same category as individuals who have sustained blunt and/or penetrating trauma, and other nuances remain debatable. In addition, the acute lung injury associated with the transfusion of blood products (TRALI), which may differ in mechanism of injury, can occur in conjunction with trauma or as a separate, unique risk factor. Similarly, drug toxicity and smoke inhalation can produce acute non-cardiogenic pulmonary edema that may be indistinct clinically from ARDS, but the pathogenesis of the lung injury may be different from, for example, sepsis. The concept that clinical risk factors could result in a similar clinical picture but the mechanism of lung injury could be variable was highlighted by a study from our group demonstrating that patients with sepsis and trauma who developed ARDS actually had different measurements of biomarkers at the time of their ARDS diagnosis (8). Other studies found significant differences in mortality based on the risk factor ranging from low mortality with trauma and a significantly higher mortality with sepsis (9). These observations led to increased discussion of the impact of patient heterogeneity on risk, pathogenesis, and outcomes from ARDS. In addition to the identified clinical conditions that increase risk for the syndrome, many other factors have been reported that impact the development and outcomes of ARDS. They include gender, race, age, genetics, the environment, microbiome, alcohol use, diabetes, kidney disease, body mass index (BMI), smoking, and cardiovascular disease (7). Our group showed early on, for example, that the incidence of ARDS is greater in patients with a history of alcohol use (10) and lower in patients with diabetes and sepsis (11). These and other studies clearly emphasized the importance of recognizing patient heterogeneity.

The issue of heterogeneity continues to be a focus in ARDS research and was the topic of the Thomas Petty Aspen Lung Conference in 2021 entitled “ARDS in the 21st Century: New Insights into Clinical and Mechanistic Heterogeneity.” The National Institutes of Health National Heart, Lung, and Blood Institute (NIH:NHLBI) just released a Request for Application to understand heterogeneity and underlying mechanisms of critical illness syndromes (https://grants.nih.gov/grants/guide/rfa-files/RFA-HL-23-001.html).

Study Design

The impact of study design has been a significant issue in biomarker studies. For studies focused on identifying diagnostic and prognostic biomarkers, two major issues were identified early on: control group and timing. For example, in an early unpublished study comparing plasma levels of IL-8 from normal healthy subjects to those of patients with ARDS, the differences between the groups was highly significant. However, when IL-8 levels were measured sequentially in the plasma of patients with sepsis who did and did not develop ARDS, there were no differences between the groups. Biomarker measurements also vary with the site of measurement. Lung biopsies in ARDS patients are rarely available so studies are limited to blood, sampling of alveolar fluid, and urine. It is not uncommon for measurements between plasma and alveolar sampling to vary significantly (12). Differences in measurement techniques also contribute to differences in results between studies. One of the biomarkers measured early on was tumor necrosis factor (TNF). TNF levels varied between investigator groups, which led to different conclusions about the potential import of that cytokine. In our lab, we compared multiple assay methods available for TNF at the time and found that the results varied widely and there was no independent standard (13), which accounted for some of the -differences in results between centers.

Identification of Phenotypes

As more biomarkers were identified over time, studies expanded from measuring single biomarkers in isolation to measuring multiple biomarkers. These included pro- and anti-inflammatory mediators, measurements of epithelial and endothelial injury, and markers of lung and other organ injury, as well as clinical biomarkers such as oxygenation. This presented a challenge for the overall analysis and interpretation of the results. Dr. Carolyn Calfee and her colleagues used risk reclassification with multiple biomarkers that include both clinical and biochemical measurements and showed that they could improve mortality prediction in acute lung injury (14). Those studies have progressed dramatically under the guidance of Dr. Calfee and her colleagues, and two very distinct phenotypes have been identified: one is hyper-inflammatory and the other is hypo-inflammatory. These two phenotypes not only vary in biomarker measurements, clinical characteristics, response to interventions, and outcomes, but they are also stable over time (7). Interestingly, they have recently been shown to vary with alcohol consumption and cigarette smoking, again emphasizing the extraordinary complexity of studying this syndrome (15). Identification of these phenotypes has helped and will continue to help us identify more homogeneous patient populations for clinical trials and studies of pathogenesis and could contribute to developing personalized approaches to therapeutic interventions.

CLINICAL TRIALS NETWORKS

The creation of the NIH:NHLBI ARDS Clinical Trials Network (ARDSNet) in 1994 expanded the size and scope of biomarker studies. This network included 10 sites and a coordinating center and ultimately enrolled more than 2,000 patients in clinical trials. Initially, the network was designed only for clinical trials with no funding for basic or translational research. A group of investigators lobbied the NHLBI to expand the scope of the network which resulted in increased funding to collect biologic samples from the patients and more carefully phenotype each subject. Almost all the patients enrolled in ARDSNet trials had well-characterized phenotypes and blood (plasma) and DNA samples. In the landmark trial demonstrating the mortality benefit from low tidal volume ventilation (16), multiple studies using the banked biologic samples demonstrated that biomarker measurements differed between those individuals ventilated with 6 ml/kg predicted body weight (PBW) versus 12 ml/kg PBW. Our group published the first of those studies showing significant differences in IL-6, IL-8, and IL-10 (17). Many other studies have followed and contributed not only to the understanding of the pathogenesis of ARDS but also to other areas including the heterogeneity in response to interventions.

THE COVID-19 PANDEMIC

The COVID-19 pandemic provided an excellent opportunity to see the progress in biomarker studies in acute lung injury. As noted, the number of patients with ARDS during the pandemic, based on modeling, is about 25 million versus the worldwide incidence of about 3 million prior to 2020. This substantial increase not only impacts the number of patients requiring clinical care, but it also significantly increases the size, scope, and number of clinical trials that can be done to further the prediction, diagnosis, and development of therapeutic interventions for these patients. Just as it has since the first recognition of the syndrome, the definition of ARDS in the era of COVID continues to evolve. Some investigators and clinicians argue that there needs to be a distinct definition of acute lung injury associated with COVID-19 (18), whereas others argue that the Berlin definition should expand to include a definition of hypoxemia that reflects the use of high flow nasal oxygen with flow rates of greater than 30 liters per minute (19), which has almost become the standard of care now for the COVID patient population and is increasingly used for all patients with ARDS.

Lessons from the past have been clearly integrated into the study of biomarkers in COVID. Control groups are generally well defined. In recent studies of cytokines in patients with COVID-19 and ARDS, for example, the control groups included patients with sepsis with and without ARDS, and patients who had sustained trauma, with and without ARDS (20, 21).

Even subsequent to the pandemic, bigger clinical trials have been supported by many large clinical trial networks around the world. Those networks were able to rapidly pivot to study COVID-19, providing a truly remarkable platform to advance science and clinical care. One example is the NIH: NHLBI Prevention and Treatment of Acute Lung Injury (PETAL) network that was established in 2013. The network, which includes 12 sites with more than 40 hospitals, led significant pivotal trials prior to the pandemic and, as with the ARDS network before it, assured that all enrolled patients had well-characterized phenotypes and banked biologic samples. As the COVID-19 pandemic rapidly evolved, the PETAL network transitioned almost immediately and launched critical clinical trials. One example that highlights the speed and complexity of those trials was ORCHID: Outcomes Related to COVID-19: a multicenter placebo-controlled study of hydroxychloroquine versus placebo. In addition to developing a rigorous study design, the investigators had to obtain an Investigational New Drug application from the Food and Drug Administration, develop methods for consent when patient and family access was limited, and create study protocols that could be implemented in an era of limited personalized protective equipment (PPE) and a workforce facing unprecedented clinical work. The Proclamation on Declaring a National Emergency Concerning the Novel Coronavirus (COVID-19) was issued on March 13, 2020 (22), and the first patient was enrolled in ORCHID on April 2, 2020. A total of 1,889 patients were screened, 1,041 were deemed eligible, and 479 were randomized. The primary outcome was clinical status, and the trial was stopped for futility on June 19, 2020. The initial results were published online on November 9, 2020 (23). This was an incredible feat and demonstrated that large complex trials in critical illness were not only very possible but could also be done under extraordinarily challenging situations. Many other networks, clinical consortia, and collaborative sites also conducted critically important trials; in aggregate, all of these groups set a precedent for not only sharing data and biologic samples from ongoing clinical intervention trials but also collating information from the rapidly evolving experience with the clinical course and care of patients. This extensive, open collaboration allowed for even more rapid, impactful contributions to the understanding of the epidemiology, clinical course, and response to potential therapeutics and established an enduring platform for future studies.

SUMMARY

Although no biomarker yet exists for the diagnosis of ARDS, studies of biomarkers have contributed to a better understanding of the epidemiology and pathogenesis of the syndrome. They provide the potential to identify specific patient populations for clinical trials and even for personalized approaches to care and therapeutic interventions in the future. Remembering the lessons learned from the past can accelerate progress in the future.

DISCUSSION

Calkins, Baltimore: Your talk was really fantastic. I remember the days when I was at Mass General where ARDS was a really a big thing. What is the current mortality of ARDS, and is the outcome getting better? What current treatments are employed?

Parsons, Burlington: If you take COVID out of the picture and look at ARDS as a whole, the mortality rate has been significantly decreasing to around 35%–45%. In clinical trials, it is a little bit lower than that. The mortality is decreasing, but it’s not entirely clear why. Low tidal volume ventilation has definitely impacted mortality, but other things have likely contributed; early treatment and diagnosis of sepsis has potentially led to some of the decrease in mortality. There is no specific treatment for it—low tidal volume ventilation is about all we have. With COVID, a lot more literature has been coming out in terms of early prone positioning and other things that will likely be implemented more than they were before, which may further help to decrease mortality.

Militoris, Indianapolis: Polly, very nice presentation. I have followed the acute kidney injury biomarker story now for about 15 years and seen no real progress in what’s happening. If we have no treatment, why pursue biomarkers? It’s kind of the yin and yang or a circular argument; until we get a therapy that empowers biomarkers to be clinically useful and profitable, there doesn’t seem to be the impetus to move forward. CMS hasn’t sponsored any biomarkers even though the FDA has in AKI. What about this in ARDS?

Parsons, Burlington: I think that’s a legitimate point. If you flip it, one of the issues is that there have been hundreds of different trials of therapeutics in ARDS, sepsis, and others that haven’t worked. Very heterogeneous patient populations have been treated. If you could use biomarkers that are being used now to actually identify specific patient populations who might be more likely to respond, a number of therapeutic interventions may well work in some smaller groups. In the old days, we gave anti-TNF to everybody, which didn’t work; some patients didn’t actually have elevated TNF. There’s a lot of push now for clinical trials using ways to identify patient populations who would be more likely to actually respond. There is a major flaw in what we have been doing—treating everybody as if they’re all the same.

Zeidel, Boston: Wonderful talk. Thank you very much. Unlike Bruce Militoris who is into acute renal failure, I intermittingly treat renal failure. ARDS is very similar to acute kidney injury—it’s massive injury of a vital organ. It’s heterogeneous and has a course, but it’s very difficult because we don’t have treatments other than supportive. However, the ARDS network has been a real example of, and a follow on to, cooperative clinical trials where we try to do better. We’re behind in renal, we’re beginning, we’re doing some of these kinds of trials, but getting those organized and moving forward is so difficult. How did the NHLBI actually get it together so that you could do all of these trials? We haven’t done as well in the kidney field.

Parsons, Burlington: The NHLBI success came from two things: (1) it started the successful asthma net earlier, which set a precedent, and (2) many of us had SCOR grants in ARDS. The second iteration of SCOR grants required a clinical component, so you went from a basic science SCOR grant to the necessity of having a clinical component. This is how I, as a just-rising fellow, junior faculty, got to be a major player in the SCOR grant at my institution. There was nobody else in the clinical space, so I was lucky. Many of us in that space realized that there was no mechanism to begin doing these larger studies. Initially, we pushed to say there needed to be some biomarker and pathogenesis studies to just collect samples together. Then it evolved, and the NIH created the clinical network. As the network evolved, we convinced the NIH to include funding for pathogenesis studies.

Zeidel, Boston: Thanks very much.