Faculty & Staff Featured News Research

New Approach to Classifying Critically Ill Patients

In a recent study published in The Lancet Respiratory Medicine, Assistant Professor of Anesthesiology, Pratik Sinha, MB ChB, PhD, and his collaborators shed light on the heterogeneity of conditions such as sepsis and acute respiratory distress syndrome (ARDS). The study’s insights may transform the understanding and treatment of clinically ill patients. 

The research stems from the recognition that the inherent heterogeneity within critical illness syndromes has led to the failure of hundreds of clinical trials. The investigative team has previously identified two distinct phenotypes within ARDS called the hypoinflammatory and hyperinflammatory phenotypes. These phenotypes exhibited diverse clinical and biological traits, and distinct responses to interventions, prompting a broader question regarding their relevance beyond ARDS.

Utilizing latent class analysis on clinical data and protein biomarkers, this recent study reveals that the hypoinflammatory and hyperinflammatory phenotypes extend beyond ARDS, and are identifiable independently in two cohorts of critically ill sepsis patients, indicating a common underlying biological signature regardless of the critical illness syndromic diagnoses. Next, they used previously validated machine learning models, that use only routinely collected clinical data, to identify the two phenotypes in two prior randomized controlled trials: PROWESS-SHOCK (activated protein C vs placebo) and VASST (vasopressin vs norepinephrine in septic shock). 

In all cohorts, the hyperinflammatory phenotype was associated with higher proinflammatory cytokines, greater organ failure and shock, and more bacteremia. Patients with the hyperinflammatory phenotype experienced more than double the mortality and prolonged stays in intensive care units compared to the hypoinflammatory phenotype. These phenotypes were consistent with prior ARDS phenotypes. Further, in PROWESS-SHOCK, the investigators observed significant treatment interaction with activated protein C and the phenotypes for mortality. Compared to placebo, activated protein C was associated with higher mortality in the hypoinflammatory and lower mortality in the hyperinflammatory phenotype. 

The two phenotypes have now been identified in over 14,000 critically ill individuals, encompassing sepsis and ARDS, including pediatric cases. The findings suggest that the hyperinflammatory phenotype may represent a universal signature of critical illness, agnostic of the patient’s syndromic diagnosis. These phenotypes disrupt conventional assumptions of prior critical care trials that have treated the two phenotypes as a uniform population. By offering a cohesive framework aligning biological and clinical characteristics, these phenotypes have the potential to enhance the design and success of future clinical trials. 

“Traditionally, for ICU categorization of syndromes like sepsis and ARDS we have relied on clinical symptoms. The problem with this approach is that these syndromes lack specificity, encompassing various diseases, pathways, and outcomes,” said Dr. Sinha. “This heterogeneity hampers testing of targeted treatments. For example, patients recruited in trials based on these syndromes, may not consistently have the biology that we seek to treat, even though they may look similar from the standpoint of their symptoms. Using our phenotyping schema, a consistent biological signature emerges, tied to divergent outcomes and varied responses to therapies.”

Dr. Sinha emphasizes the necessity of further prospective trials tailored to these phenotypes before their clinical integration. He explains that the hyperinflammatory phenotype denotes a distinct biological pattern linked to worse outcomes that may be potentially modifiable. It presents a treatable trait that has a quantifiable diagnostic signature, with potential to reshape critical care. While these findings hint at biological precision in the ICU, their true impact relies on practical demonstration of treatment responsiveness in future phenotype-informed trials. Further research will delve into why some patients manifest with the hyperinflammatory and others with the hypoinflammatory signature with seemingly the same insult.

Dr. Sinha’s study marks a significant stride toward personalized critical care. By better characterizing these phenotypes’ biological traits, new therapies can be identified for phenotype-specific trials, which may significantly improve intervention effectiveness and patient outcomes in future studies.