VIDEO: Emerging therapies in sepsis management

Editor’s note: This is an automatically generated transcript, which has been slightly edited for clarity. Please notify editor@healio.com if there are concerns regarding accuracy of the transcription.

I think the most promising therapies are those that focus on modulating the immune system. Since sepsis really is a disorder that involves a dysregulated immune system. So examples include agents that inhibit the innate immune system, or perhaps inactivate or remove harmful cytokines. However, no such agents are really close to FDA approval or even showing clear efficacy in clinical trials as of now, and many sepsis therapies that have looked promising based on animal studies or small single center studies have not panned out when tested in humans or in larger randomized controlled trials. I think the latest example of this is vitamin C with thiamine and hydrocortisone, which was promising in a single center retrospective study but subsequently produced disappointing results in larger randomized controlled trials.

Now, one major reason why a lot of potential therapies have not panned out, I think, is because the patient population that has sepsis, in general and in specific trials, have been very heterogeneous. Sepsis includes a broad range of infections, organ dysfunctions and patients with different host factors, as well as different immunological pathways. So many of us believe that we need to do a better job of selecting patients for trials to match them to the likely action of the drug in question. But this, in turn, requires better biomarkers of sepsis and the specific immune response.

The last thing I'll say on this topic is that while not specifically therapy per se, there's a lot of interest in research going on with using machine learning and artificial intelligence to either diagnose sepsis or to predict its onset several hours in advance using clinical data that are available in the electronic health record. Several studies have suggested that these types of algorithms can be very accurate, but we definitely need more studies assessing the real-life impact of machine learning-based alerts on patient outcomes.