Big Data and hidden subtypes of sepsis

Results of a study conducted by the University of Pittsburgh School of Medicine suggests that sepsis is not one condition, but many conditions that could benefit from different treatments. The findings are published in JAMA and were presented at the American Thoracic Society's Annual Meeting.

Sepsis is the number one killer of hospitalised patients and is a life-threatening condition that arises when the body's response to infection begins to injure its own tissues and organs. It has been over a decade, and no major breakthroughs have happened in the treatment of sepsis. The only improvement observed so far is the enforcement of the "one-size-fits-all" approach for prompt treatment, highlights Christopher Seymour, associate professor in Pitt's Department of Critical Care Medicine and member of Pitt's Clinical Research Investigation and Systems Modeling of Acute Illness Center. But as he explains, these protocols ignore the fact that all sepsis patients are not the same. It is a condition that kills nearly 6 million people annually, and using a one-size-fits-all approach is unacceptable for such a huge threat to patients.

By seeing sepsis as several different conditions, and with varying clinical characteristics, it may be possible to discover and test therapies that are tailored to treat the different subtypes of sepsis.

The “Sepsis ENdotyping in Emergency Care” (SENECA) project, funded by the National Institutes of Health (NIH), has used computer algorithms to analyse 29 clinical variables found in the electronic health records of more than 20,000 patients who had sepsis within six hours of hospital arrival. Patients were clustered into four distinct sepsis types, which include:

1. Alpha - found to be the most common (33%) and with the least organ dysfunction and lowest in-hospital death rate at 2%.
2. Beta - found in approximately 27% of patients; mostly elderly patients with the most chronic illness and kidney dysfunction.
3. Gamma - almost the same frequency as beta; but associated with greater inflammation and pulmonary dysfunction.
4. Delta - least common at around 13%; most deadly type, often associated with liver dysfunction and shock; showed highest in-hospital death rate at 32%.

After studying another 43,000 sepsis patients, the UPMC team confirmed these findings. These findings were then applied to recently completed international clinical trials that tested different therapies for sepsis. None of these trials had anything significant to report. However, results might have been different if the trial participants had been classified on the basis of these four subtypes. For example, early goal-directed therapy (EGDT) was not found to have any benefit following a five-year study, but when the UPMC team re-examined the results, they found that it could have benefitted the Alpha type of sepsis patients but would result in worse outcomes for the Delta subtype.

If you think about it logically, the theory of sepsis subtypes makes perfect sense. Just like all breast cancer patients do not receive the same treatment (as some breast cancers can be more invasive and require aggressive treatment), the same is true for sepsis. There is thus a need to find therapies that apply to specific types of sepsis and then design clinical trials to test those therapies.