A Prospective Machine Learning Model for the Identification of Sepsis Endotypes

Lead Investigator: Ritu Lal, GEn1E Lifesciences Inc.
Title of Proposal Research: A Prospective Machine Learning Model for the Identification of Sepsis Endotypes
Vivli Data Request: 9388
Funding Source: Funding by GEn1E Lifesciences
Potential Conflicts of Interest: Statement of No Additional Conflicts:
We hereby affirm that beyond their respective roles and affiliations with GEn1E Lifesciences Inc. (the funding source of the study), there are no additional financial or non-financial conflicts of interest for any of the aforementioned team members within the last three years.
Management of Conflicts of Interest:
To manage the disclosed conflicts of interest, namely the affiliations with GEn1E Lifesciences Inc., we commit to full transparency. All aforementioned conflicts of interest will be fully disclosed when the research associated with this data request is presented or published. This ensures that all stakeholders are aware of these affiliations, upholding the integrity and impartiality of our research.
We understand the importance of transparency and ethical conduct in research and are committed to adhering to these principles throughout our involvement in this project.

Summary of the Proposed Research:

Sepsis is a leading cause of death in serious COVID patients with no FDA-approved therapeutic treatment and has a 40%+ mortality rate. Current research in the field of Sepsis indicates that there is heterogeneity (i.e., differences) in the treatment effects (HTE) amongst Sepsis patients. This HTE is one of the major reasons why numerous trials have failed to deliver effective therapies.

Previous studies have only retrospectively identified the HTE amongst patients, but have not looked at this heterogeneity proactively. We have a differentiated approach where we intend to develop an machine learning-based model for the proactive determination of endotypes (sub-categories) of Sepsis patients. A proactive approach can be helpful to us through predictive enrichment, which seeks to identify patients that are most likely to benefit from a novel therapy. Predictive enrichment means that instead of treating all patients with sepsis in the same way, enrichment strategies aim to separate the patients according to the different factors that may be contributing to their disease.

Biomarkers are biological molecules found in blood, other bodily fluids, or tissue which represents a sign of a normal or abnormal process in the body. As a first step, we will potentially explore developing a composite score (a numerical score made up from individual scores assigned to each factor of interest) based on biomarkers and clinical variables measured in the emergency room (ER) and other point-of-care outcomes to distinguish between the two endotypes.

Through access to anonymized and de-identified patient data from the study, we will use statistical and machine learning methods to proactively determine endotypes of Sepsis patients. Confidentiality and privacy will be maintained by storing the study data in a secure, encrypted, and password-protected analysis environment at GEn1E Lifesciences. We also understand that the data will be provided in the secure Vivli research environment and we agree and abide by the data sharing policies and anonymization standards set by the data contributor. This proposed research does not involve the use of any other data sources other than the study data.

Requested Studies:

A Randomized, Double-blind, Placebo-controlled, Multicenter, Phase 3 Study of Drotrecogin Alfa (Activated) Administered as a Continuous 96-hr Infusion to Adult Patients With Septic Shock
Data Contributor: Lilly
Study ID: NCT00604214
Sponsor ID: 11940