Combining metabolomics and proteomics, a consortium of academic and industry researchers has identified a signature that appears useful for predicting death in sepsis patients.
Consisting of five metabolites as well as age and hematocrit, the profile distinguished between sepsis survivors and non-survivors with area under the curve of 0.74 in a study of 86 sepsis patients published this week in Science Translational Medicine.
According to Raymond Langley, a researcher at Lovelace Respiratory Research Institute and first author on the paper, the group has filed patents covering the signature and hopes in the future to license it to an industry partner with the ultimate goal of incorporating it into a point-of-care device that can be used to assess sepsis patients' risk of death.
Sepsis is the tenth leading cause of death in the US, with sepsis infections costing the country's healthcare system an estimated $16.7 billion per year.
"There are about 11 million patients a year who wander into a doctor's office or emergency department with signs of sepsis, Stephen Kingsmore, a researcher at University of Missouri-Kansas City School of Medicine and senior author on the STM paper, told ProteoMonitor. "You can't admit 11 million patients into the hospital, though. So the vast majority of these patients are prescribed rest and an oral antibiotic ... [while] typically the patients who get admitted are the ones who are really sick or have other co-morbidities [like] diabetes or renal failure."
This approach, however, misses a subset of patients who ultimately develop severe sepsis and go on to die, Kingsmore said. "So the utility of this test would be as a kind of simple screen to do in the 11 million patients to pick out the ones who are not going to do well and get them into the hospital."
Reinforcing the need for such a test is the researchers' finding that while their signature could predict patient survival, survival did not appear to be related to the severity of the patient's condition at the time of admission.
"Going into this, we had this classic idea of sepsis progression," Kingsmore said. "You got an infection, and then you got sepsis, and then you had severe sepsis, and then you got septic shock and died."
In fact, however, the researchers found that regardless of what stage of sepsis patients were in, the metabolite signature separated them into survivors and non-survivors.
"From the moment a patient arrived in the emergency department, they had this pre-programmed response," Kingsmore said. "It didn't matter what stage of sepsis they were in, this biochemical signature binned people into the two groups."
This result was "completely unexpected," he said. However, he noted that recently other researchers have made similar observations in other areas, suggesting, perhaps, a more general phenomenon at work.
For instance, "people have now had datasets that are yielding the same kind of results in trauma," he said. "Which is not that different, if you think about it – a patient has an acute illness and some survive and some don't."
The study emerged from antibody chip work profiling late-stage sepsis patients that Kingsmore did roughly a decade ago in collaboration with Eli Lilly.
"We found what we thought were very compelling differences among patients who had sepsis and didn't," he said. "And so we used that data to say, 'Let's move earlier in sepsis and see if we can pick up signatures that differentiate folks who are infected from those who aren't.'"
That work, Kingsmore noted, was primarily proteomic-focused until Pfizer offered the researchers funding to perform metabolomic profiling of these patients. "And it turned out that that was where the paydirt was," he said. Pfizer, he noted, has waived its rights to any intellectual property stemming from the research.
In the study, the researchers performed proteomic and metabolomic profiling of a 150 patient discovery set, followed by validation in two addition cohorts – one a set of 18 sepsis nonsurvivors and 34 matched sepsis survivors and the other a set of 29 noninfected patients with systemic inflammatory response syndrome, 36 sepsis survivors, and 25 sepsis nonsurvivors.
Via this analysis, they identified 82 metabolites and 59 plasma proteins that differed significantly between survivors and non-survivors. Working under the notion that true changes in the metabolome would likely be accompanied by analogous changes in the proteome, they integrated these two datasets by performing global cross-correlation and hierarchal clustering of matched metabolites and proteins. In this way, they were able to identify a number of unannotated peptides and metabolites and assign them into families as well as identify regulators of metabolic pathways, Kingsmore said.
Ultimately, however, the proteomic data "was fairly disappointing," he said, noting that the mass spec platform they used – a Thermo Fisher Scientific LTQ linear ion trap – was not sensitive enough to pick up low abundance proteins, particularly cytokines, where, Kingsmore said, "most of the action is in sepsis."
The researchers built their signature, therefore, based on the metabolomic data. Langley said that he and his colleagues may take up the proteomics angle again in the future but that for now they are focused on the metabolomics.
He added that he has since repeated the study in animal models including non-human primates, and discovered that the human findings "almost perfectly translate" to the models.
Langley is also working to integrate transcriptomic data with the metabolomic information, an effort that he said has allowed him "to really refine the pathway [information] and see what pathways are being effected and how they are being affected, which I think gives us the potential for some future therapeutic studies."
He noted that while he ultimately hopes to license the signature to a firm capable of developing it for a point-of-care device, "we are still early in that process."
As a significant cause of patient deaths and driver of healthcare costs, sepsis has received considerable interest from the testing industry. For instance, on the protein front, Thermo Fisher offers its procalcitonin assay for the diagnosis of the condition. Protein biomarker firms Pronota and Astute Medical also have proteomic tests for sepsis in their development pipelines.
Nucleic acid-based approaches are also under development, with firms like Cepheid and BioMérieux, for instance, working on PCR-based tests for identifying the infectious agents responsible for the condition.