SCCM Resource Library-What Can Biomarkers Tell Us?

Video Transcription

Thank you to the Society of Critical Care Medicine and the organizers of this session on pediatric ARDS, a translational reflection.  I am Jocelyn Grunewald, an assistant professor of pediatrics at Emory University and Children's Healthcare of Atlanta.  Today I will be speaking about what biomarkers can tell us about pediatric ARDS. First, I think we should consider the overarching framework of this topic.  ARDS is a heterogeneous syndrome characterized by severe refractory hypoxemia with acute onset of diffuse bilateral pulmonary edema that's not fully explained by cardiac dysfunction  or fluid overload. I think of ARDS as an umbrella term under which there are clinically observable characteristics  or phenotypes that contribute to clinical presentation, physiologic responses, symptoms, triggers, and exposures.  We're at the stage of defining different subgroups or phenotypes in pediatric ARDS,  but we lack understanding of how these phenotypes are linked to biological mechanisms and molecular pathways, otherwise known as endotypes.  In addition, there is variability and uncertainty in the sample size needed to identify phenotypes.  We do not know the optimal frequency and timing of measurement or distribution of data collection procedures. Statistical modeling must also be considered.  The ultimate goal of biomarkers is to help us identify phenotypes and endotypes of children with ARDS to deliver personalized therapies.  The problem with refining pediatric ARDS using biomarkers is threefold. First, ARDS has high clinical heterogeneity. Second, there's poor recognition of ARDS.  Third, there is poor correlation of an ARDS diagnosis with diffuse alveolar pathology. The goals of biomarkers are to reduce heterogeneity and increase specificity.  Today, biomarkers are not prospectively used to diagnose, prognosticate, or predict response  to supportive therapies in ARDS because biomarker use has not yet been clinically demonstrated. So what is a biomarker?  A biomarker may be a physiologic characteristic or a biological molecule found in blood or  body fluids that is a sign of a normal or an abnormal process or of a condition or disease. Biomarkers have a variety of purposes.  They can be used to stratify susceptibility or risk. They can be used as a diagnostic tool. And they can be used for prognostic or predictive purposes.  We are often most interested in using biomarkers for prognostic and predictive enrichment. For example, from a group of critically ill children with ARDS, we can first make two  subgroups of children. The top group is at low risk of death. The bottom group is at high risk of death.  Using severity of hypoxemia is one way to perform prognostic enrichment of children with a high risk of death.  The children with mild ARDS who are at low risk of death would receive standard care. The children with severe ARDS at high risk of death would be sorted into endotypes using  biomarkers that link to an underlying mechanism. Biomarkers linked to an endotype or biological pathway result in predictive enrichment.  The ultimate goal is that predictive enrichment leads to targeted or personalized therapies for ARDS.  There are several things to consider when using biomarkers for diagnostic, prognostic, or predictive enrichment.  First, when testing a biomarker, the choice of a control or comparison group is key.  Even in critical care, we are trying to differentiate between a sick child and a really sick child.  Severity is a matter of degrees and is complicated by different comorbidities, triggers, and exposures. Second, timing of sampling is important.  We need to consider the duration and stability of the biomarker. We tend to think of ARDS as occurring in linear phases, but this is likely not the case.  Finally, which compartment should be sampled? Blood or plasma is an accessible measure of systemic inflammation, however biomarkers  in the systemic circulation may not reflect the pathobiology occurring in the lungs. In this case, sampling airway fluid may be a better marker of local inflammation, especially  if a pulmonary process rather than a systemic process triggered ARDS. Biomarkers come in many categories ranging from inflammatory cytokines to metabolites,  Transcriptomics using microarrays and RNA-seq techniques have been used to discover the differential gene expression signature of ARDS.  Finally, functional immunophenotyping such as analysis of neutrophil activation and neutrophil  extracellular trap release or cellular response to a stimulus have been studied in ARDS. There are many examples of single or small combinations of plasma prognostic biomarker  candidates in ARDS. Examples of biomarkers categorized by their relationship with the endothelium, lung epithelium, inflammation, and coagulation processes have been reviewed.  Three recent reviews are referenced with citations to many of the primary articles. Our adult colleagues were the first to use more sophisticated modeling techniques such  as unsupervised clustering and latent class analysis to discover biomarker panels. These studies were conducted in the context of a rich biobank of samples from patients  enrolled in large randomized control clinical trials. Two subphenotypes were discovered, a hyperinflammatory phenotype characterized by higher mortality  and fewer ventilator-free days in comparison to a hypoinflammatory phenotype. In post-hoc analyses, participants with these two phenotype subgroups respond differently  to randomized fluid management and PEEP. The hyperinflammatory phenotype has, on average, higher levels of IL-6, IL-8, and TNF receptor-1  and lower levels of bicarb and activated protein C compared to hypoinflammatory phenotypes. This methodology has now been repeated in a pediatric cohort with children enrolled  in the RESTORE trial. RESTORE was a study that randomized children with acute hypoxemic respiratory failure to nurse-led protocolized versus non-protocolized sedation.  The Biomarkers in Acute Lung Injury, or BALI, cohort was an ancillary study that collected biological samples from children enrolled in RESTORE.  Using the same latent class analysis methods as the groups led by Drs. Kalfi and Maffei, the BALI investigators similarly found two subphenotypes of children.  As in the adult study, there was a hyperinflammatory group characterized by high soluble TNF receptor-1, IL-6, and IL-8 compared to the hypoinflammatory group.  The top eight biomarkers are circled in green and listed in the green box in the middle of the screen.  As shown in the bar graph on the right side of the slide, a higher proportion of children in the hyperinflammatory group, shown by the orange bar, were categorized as meeting the  criteria for severe ARDS. Although not shown in the slide, children with the hyperinflammatory phenotype had on  average a longer duration of mechanical ventilation, 10 versus 6 days, and a higher incidence of  mortality, 14 versus 2 percent, compared to children with the hypoinflammatory phenotype.  A parsimonious model akin to the adult findings showed that when using only three biomarkers,  soluble TNF receptor-1, IL-8, and use of vasopressors, that the discriminatory ability of this panel  to distinguish between hyperinflammatory and hypoinflammatory phenotypes was excellent, with an area under the receiver operating curve of 0.95.  One of the first pediatric latent class analyses was performed in 235 children with ARDS using a cohort comprised of children recruited from five hospitals over a 10-year period  from 2008 to 2017. Plasma samples were collected on day one of a diagnosis of ARDS. Matrix metalloproteins and their endogenous inhibitors were measured and associated with  biomarkers of inflammation, endothelial injury, impaired oxygenation, and outcomes of morbidity and mortality.  The MMPs were classified into two group profiles, as shown in the leftmost graph with black and gray bars denoting the two MMP subgroups.  The center survival graph shows the results from a geographically distinct validation cohort.  In the validation cohort, approximately one-third of patients were in the MMP Profile 1 group, which conferred a fourfold increased odds of death or severe morbidity when adjusting  for severity of hypoxemia using the PF ratio and for other confounders. Multivariable logistic regression was used to model the prognostic discrimination of  severe morbidity and mortality using the MMP subgroup. There was a modest discriminatory ability of the model to predict mortality and severe  morbidity, with an area under the receiver operating curve of approximately 0.75. Another study used classification and regression tree modeling to adapt, or re-derive, a plasma  biomarker-based sepsis mortality risk stratification tool for pediatric ARDS. This model was called PARDS-SEVERE.  The 152 children in this cohort were split into a derivation and a validation set, and three biomarkers, CCL3, HSP, A1b, and IL-8 plus age, were used to predict 28-day mortality  with a good area under the receiver operating curve of 0.85. In a separate study of whole blood transcriptomic profiles of 96 children with ARDS, three clusters  or subphenotypes were discovered using k-means clustering. The three clusters are shown in the cluster map in the top graph.  Children in the three subphenotypes had different clinical characteristics. Children in phenotype 1, shown in red, had persistent hypoxemia and the highest mortality at 32%.  Children in phenotype 2, shown in green, had the highest proportion of immunocompromised patients, had rapidly resolving hypoxemia, and 24% mortality.  Children in phenotype 3, shown in blue, had the fewest comorbidities and 8% mortality.  Phenotype 3 was associated with lower mortality and a higher probability of extubation relative to phenotype 1 after adjusting for multiple confounders.  The results are not shown on the slide, but upregulated pathways for phenotype 1 involved T-cell receptor, PD-1 receptor signaling, and Th1, Th2, and Th17 differentiation.  These findings need to be validated with an external cohort. There is no known systemic gene expression signature of ARDS in children.  Dr. Rishi Kalameswaran and I collaborated with Drs. Ye Ya and Wang to use their publicly available gene expression data to help address this knowledge gap.  To do this study, we compared children with a PF ratio under 200 to those with a PF ratio greater than or equal to 200.  The dataset GSE 147902 contains microarray data using whole blood from children who met the Berlin definition of ARDS.  The dataset GSE 66099 contains data using whole blood from children with sepsis-related acute hypoxemic respiratory failure.  We ranked the top 1,500 differentially expressed genes in each dataset and serially tested the addition of the common ranked genes to a model discriminating children by severity  of hypoxemia. The center panel graph shows the receiver operating curve characteristic with the addition of genes to the model. The red line shows the Berlin ARDS data.  The black line shows the sepsis-acute hypoxemic respiratory failure dataset. The ROC characteristic plateaus around 25 to 30 genes for both datasets.  And as expected, the model performs better in the Berlin ARDS-defined dataset, shown by the higher ROC values for the red line compared to the black line in the center graph.  Next, we used the top 185 genes common to both datasets to perform a CAG pathway analysis.  Metabolic pathways were the most overrepresented in the ARDS gene expression signature analysis.  Work in my lab is focused on assessing the airway immune response in intubated children with ARDS.  In collaboration with Dr. Rishi Kalamaswaran's biomedical informatics group at Emory University,  we have used machine learning algorithms to discover a differential gene expression signature from airway cells obtained from tracheal aspirate samples from children with ARDS.  We have also performed a cluster analysis and profiled metabolites from the airway fluid of intubated children with acute hypoxemic respiratory failure.  We found three clusters of metabolites from the airway fluid in these children. Children within these clusters had different baseline characteristics and duration of mechanical  ventilation. Finally, we have studied airway neutrophil activation markers, the type I interference  signaling pathway, and neutrophil extracellular trap release in the airway fluid from children with ARDS.  These markers are different between children with and without ARDS, and the markers are also associated with having fewer ventilator-free days.  So what are the bottlenecks in the use of biomarkers at the bedside? First, we lack high-throughput, human-like preclinical models for efficient hypothesis  testing of molecular mechanisms of ARDS. These preclinical models could be adapted for in vitro clinical sample testing to link phenotypes with endotypes.  These models could also be used as a candidate drug screening and testing platform. Second, we lack rapid and reliable bedside diagnostics of validated biomarkers for the  diagnosis, prognostication, and prediction purposes. Point-of-care tests are ideal, but clinical laboratory tests with a rapid turnaround time are acceptable.  However, the results need to be interpretable readily and actionable at the bedside in order to be useful.  An example of a candidate human-like preclinical model is shown and has been developed in collaboration  with my lab and Dr. Shuichi Takayama's lab at the Georgia Institute of Technology. This model uses cultured human endothelial and small airway epithelial cells grown at  an air-liquid interface to mimic an alveolar capillary unit. Neutrophils can be isolated from a human donor and are allowed to transmigrate from  the upper blood side of the model through the endothelial epithelial layers into the lower airway side of the model.  The neutrophils are attracted to patient airway fluid in the media in the lower chamber. We plan to use this model for hypothesis-driven mechanistic discovery and drug testing.  I will close with a few comments about what I think is needed to implement biomarkers into routine bedside care.  A goal of biomarker discovery is the delivery of individualized or personalized care. This is a change in mindset from the last several decades of critical care, which has  focused on the development and implementation of protocolized care bundles. The goal of protocolized care is to deliver the most effective and supportive therapies  in the timeliest manner as part of care bundles. Successful implementation of biomarkers into critical care requires us to adapt our thinking  because the promise of biomarkers to deliver timely care in a targeted and tailored manner to individuals rather than entire groups of patients.  Finally, the successful discovery and clinical implementation of biomarkers in ARDS requires a change in how we perform science.  Individual scientists do not possess the expertise in all areas needed to make biomarkers a bedside reality. Therefore, teams of scientists are needed.  Team members will need expertise in basic and translational science. Teams will need biomedical engineers for model development and assay design.  Data scientists and statisticians are needed to incorporate clinical and physiologic data into diagnostic, prognostic, and prediction models.  Finally, clinical and implementation scientists are needed to study how to deploy biomarker technology to the bedside to make biomarker use in ARDS a practical reality. Thank you.  I'm happy to take any questions as part of the panel discussion.

Video Summary

In this video, Dr. Jocelyn Grunewald talks about the use of biomarkers in pediatric ARDS (Acute Respiratory Distress Syndrome). ARDS is a syndrome characterized by severe hypoxemia with diffuse bilateral pulmonary edema. Biomarkers are physiological characteristics or biological molecules found in blood or body fluids that can indicate a normal or abnormal process or condition. They can be used for risk stratification, diagnosis, prognosis, and predictive purposes. In pediatric ARDS, biomarkers can help identify different subgroups or phenotypes and link them to biological mechanisms and molecular pathways. However, there are challenges in using biomarkers, including clinical heterogeneity, poor recognition of ARDS, and poor correlation with alveolar pathology. Current research is focusing on identifying biomarker panels that can accurately distinguish between different phenotypes and predict outcomes in pediatric ARDS, but more research and collaboration among scientists with different expertise are needed for successful clinical implementation of biomarkers.

Asset Subtitle

Pulmonary, Pediatrics, 2022

Asset Caption

Pediatric acute respiratory distress syndrome (PARDS) is a manifestation of severe lung injury with a mortality rate of up to 35%. Despite this significance of PARDS in critically ill mechanically ventilated children, respiratory management remains largely supportive, with no data to support one approach over another. This session will highlight advances in the understanding of PARDS by relating experimental models to clinical experience and discussing recent advances in respiratory monitoring and management of PARDS and the identification of biomarkers that can be used for disease stratification and prognostication. Speakers will also discuss how to design clinical trials of PARDS, delivering much-needed scientific knowledge

Meta Tag

Content Type Presentation

Knowledge Area Pulmonary

Knowledge Area Pediatrics

Knowledge Level Intermediate

Knowledge Level Advanced

Membership Level Select

Tag Acute Respiratory Distress Syndrome ARDS

Tag Pediatrics

Year 2022

Keywords

biomarkers

pediatric ARDS

risk stratification

phenotypes

molecular pathways

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.