Zachary Aldewereld, MD,
Assistant Professor, Pediatric Critical Care Medicine and Infectious Diseases
UPMC - Children's Hospital of PIttsburgh
Pittsburgh, Pennsylvania
Disclosure information not submitted.
Li Ang Zhang, PhD
Information Scientist; Professor
Pardee RAND Graduate School, United States
Disclosure information not submitted.
Robert Parker, PhD
Associate Dean for Graduate Education, Professor
Swanson School of Engineering, United States
Disclosure information not submitted.
David Swigon, PhD
Associate Professor of Mathematics
University of Pittsburgh, United States
Disclosure information not submitted.
Ipsita Banerjee, PhD
Associate Professor
University of Pittsburgh, United States
Disclosure information not submitted.
Hernando Gomez, MD, MPH
Assistant Professor of Critical Care Medicine and Emergency Medicine
UPMC Prespyterian, United States
Disclosure information not submitted.
Gilles Clermont, MD, MS
Professor CCM, Mathematics, Clinical Translational Sci and IE
VA Pittsburgh Health System
Pittsburgh, Pennsylvania
Disclosure information not submitted.
Title: Identification of Severe Sepsis Clinical Phenotypes
Introduction:
Targeted therapies for sepsis, a dysregulated immune response to infection, have failed to show benefit due to a high degree of variability among subjects. We sought to demonstrate the existence of severe sepsis phenotypes that are characterized by differing clinical features and outcomes.
Methods:
A retrospective analysis was performed in a 1,023-patient cohort with severe sepsis from the ProCESS trial. Hospitals from this cohort were randomized to derivation and validation cohorts in a roughly 2 to 1 fashion, resulting in 642 subjects from 20 hospitals in the derivation cohort and 381 subjects from 11 hospitals in the validation cohort. 23 clinical variables at baseline were analyzed using hierarchical clustering. In order to remove much of the subjectivity that is inherent in determining cluster divisions in hierarchical clustering, consensus clustering, essentially a bootstrap technique utilizing repeated subsamples on which the analysis is performed, was used to identify the ideal number of clusters and validate cluster membership in the derivation cohort. Cluster robustness was then further confirmed in the validation cohort. Clusters were visualized using heatmaps over 0, 6, 24, and 72 hours. Clinical outcomes were 14-day all-cause mortality and organ failure pattern.
Results:
Five phenotypes were identified, each with unique organ failure patterns that persisted in time. By enrollment criteria, all patients had septic shock. In addition, the two high risk phenotypes were characterized by distinct multi-organ failure patterns and cytokine signatures, with the highest mortality group characterized most notably by liver dysfunction and coagulopathy while the other group exhibited primarily respiratory failure, neurologic dysfunction, and renal dysfunction. The moderate risk phenotype was that of respiratory failure, while the low-risk phenotypes did not have a high degree of additional organ failure and were distinguished by their need or lack thereof for vasopressors.
Conclusions:
Sepsis phenotypes may assist in developing targeted sepsis therapies and improving patient outcomes by providing an opportunity for early clinical actionability.