Bedside Classification of ARDS Subtype May Help Spot Patients With Higher Death Risk

CHICAGO — Classifying inflammatory subphenotypes of acute respiratory distress syndrome (ARDS) with bedside lab testing was feasible and showed differences in 60-day mortality, according to a prospective cohort study.

Among the 88 patients with a hyperinflammatory subphenotype, 60-day mortality was 51% compared with 28% in the 398 patients with a hypoinflammatory subphenotype (risk ratio 1.8, 95% CI 1.4-2.4, P<0.0001), reported Kiran Reddy, MD, of Queen's University Belfast in Ireland, at the Society of Critical Care Medicine annual meeting.

After adjustment, hyperinflammatory patients had increased odds of 60-day mortality (adjusted OR 2.7, 95% CI 1.6-4.4, P=0.0002), Reddy and colleagues wrote in their study, which was also published in Lancet Respiratory Medicine.

“This was the first prospective demonstration that you can identify these phenotypes at the bedside using rapid assay,” Reddy told attendees. That classification revealed that the hyperinflammatory group had “more sepsis, more pancreatitis or organ failures, higher APACHE [Acute Physiology and Chronic Health Evaluation] II scores, higher SOFA [Sequential Organ Failure Assessment] scores, and a lot more use of renal replacement therapy despite very similar lung injury severity score,” he said.

The Phenotypes in the Acute Respiratory Distress Syndrome (PHIND) study “shows us for the first time that we can prospectively, feasibly identify subphenotypes of ARDS in a real-world setting, and that sets us up for doing ARDS precision medicine trials,” Reddy noted.

Past studies had identified the two biologically distinct hyperinflammatory and hypoinflammatory subphenotypes of ARDS and differences in outcomes, but those studies were retrospective and had not attempted to prospectively classify patients during care.

Javier A. Neyra, MD, of the University of Alabama at Birmingham, told MedPage Today that being able to identify phenotypes of patients who might respond to specific treatments gives physicians an advantage.

“This study is a great example of how to advance precision medicine,” Neyra said, adding that it is important to support studies like PHIND that provide the predictive information needed for subtyping.

In an invited comment, Patricia Rocco, MD, PhD, of the Federal University of Rio De Janeiro, and colleagues wrote that the PHIND study “represents a methodological advance” by removing a “crucial technical barrier that has constrained progress in ARDS research” and showing “that prospective biological stratification retains prognostic relevance in practice,” despite not being designed to show potential for improving outcomes.

“Identifying biological subphenotypes is a necessary but insufficient step towards precision medicine,” they noted. “Benefit depends on whether such stratification can guide interventions that improve patient-centered outcomes.”

A barrier to such prospective classification was that models used to identify subphenotypes in previous studies were “trained to predict on standardized values of the input variables” and required “large datasets and knowledge of the population-level mean for each input variable,” the authors wrote.

They therefore developed several parsimonious logistic regression models that used three or four biomarkers and clinical variables to identify subphenotypes with high accuracy, with an area under the curve (AUC) of 0.92 to 0.96. The possible inputs included interleukin (IL)-6, IL-8, protein C, soluble tumor necrosis factor receptor-1 (TNFR1), bicarbonate, and/or vasopressor use.

They then designed the current study to quantify the biomarkers and determine whether one of the logistic regression models could be used to identify subphenotypes.

From November 2019 through September 2023, 1,853 patients from 30 intensive care units (ICUs) in the U.K. and Ireland were screened for eligibility, and 512 adults with ARDS or acute hypoxemic respiratory failure were included within 72 hours of onset of the syndrome. All patients were receiving invasive or non-invasive mechanical ventilation or high-flow nasal oxygen. Mean age was 57, 60% were men, 87% were white, 4% were Black, and 3% were Asian.

Blood samples were collected at enrollment from an indwelling arterial or central venous line and tested for IL-6, TNFR1, and arterial bicarbonate. They calibrated and used a multiSTAT device to assess IL-6 and TNFR1 within 1 hour of blood collection to use in the subphenotyping model to classify patients.

Patients were classified as hyperinflammatory if their probability of that subphenotype was greater than 50% using a model whose AUC was 0.94. Among the 490 patients who could be subphenotyped on the basis of those labs, 18% were classified as hyperinflammatory and 82% were classified as hypoinflammatory. After four patients withdrew consent, the researchers compared 60-day mortality in the remaining patients.

The 28-day mortality rate was also significantly different between the hyperinflammatory (44%) and hypoinflammatory (23%) subtypes (P<0.0001). Hyperinflammatory cases had a median of zero ventilator-free days at day 28, compared with a median 14 days for hypoinflammatory cases (P<0.0001).

Secondary outcomes that did not significantly differ between the subtypes included re-intubation rate, duration of days with ventilation, length of ICU stay, and length of hospital stay. When the last three outcomes were separated between survivors and non-survivors, however, they did differ significantly between the subphenotypes.