鶹ýӰ

Study Pinpoints 4 Novel Sepsis Phenotypes

— Distinguishing different subtypes of sepsis sets the stage for individualized treatment, researchers say

MedpageToday

DALLAS -- Four clinical sepsis phenotypes were identified that correlated with host-response patterns and clinical outcomes, researchers reported here.

The four novel sepsis phenotypes -- alpha (α), beta (β), gamma (γ), and delta (δ) -- with different demographics, laboratory values, and patterns of organ dysfunction were derived, validated, and shown to correlate with biomarkers and mortality, according to Christopher Seymour, MD, of the University of Pittsburgh School of Medicine, and colleagues.

Their analysis also identified a subtype of patients at high risk for liver dysfunction and shock who had the highest in-hospital death rates, they wrote in . The results were presented simultaneously at the American Thoracic Society annual meeting.

When the researchers revisited several high-profile clinical trials that failed to show a benefit for investigational sepsis treatments, they concluded that poor responses in the single subgroup may have led to the negative findings.

Seymour told MedPage Today said this kind of research "is starting to unpack the different subtypes of sepsis and setting the stage for individualized treatment."

"Right now, our treatment approach to sepsis is basically 'one size fits all,' whether you are a 40-year-old with influenza complicated by [a] staff infection or an 80-year-old with multiple comorbidities and biliary sepsis," he said, adding that international sepsis practice guidelines recommend the same bundle of care for everyone.

He said most previous studies searching for meaningful differences among sepsis patients have involved the measurement of gene expression, which is costly, time consuming, and not readily adaptable to the bedside.

"Ours is the first study to use electronic health record [EHR] data in a population of tens of thousands on arrival at the hospital," he said. "The data is in electronic databases that is readily adaptable to health systems."

Study Details

Seymour and colleagues examined 29 specific variables recorded in the EHR system of >20,000 patients identified with sepsis at 12 Pennsylvania hospitals from 2010 to 2012 (derivation cohort).

Reproducibility of the findings was measured in a second dataset involving around 43,000 total patient arrivals at the Pennsylvania hospitals from 2013 to 2014 (validation cohort).

The derivation cohort included patients (mean age 64; 50% men) with a mean maximum 24-hour Sequential Organ Failure Assessment (SOFA) score 3.9. The validation cohort included patients (mean age 67; 51% men) with a mean maximum 24-hour SOFA score of 3.6.

The authors reported the following on the four distinct sepsis subtypes:

  • alpha (α): Identified in one of three patients; characterized by having the fewest abnormal laboratory test results, the least organ dysfunction, and the lowest in-hospital death rate (23%)
  • beta (β): Identified in 27% of patients; defining characteristics including older age, more chronic illness, and kidney dysfunction
  • gamma (γ): Identified in roughly one in four patients; distinguished from β by having elevated inflammation and primary pulmonary dysfunction
  • delta (δ): Identified in 13% of patients; least common and most deadly phenotype; characterized by liver dysfunction and shock and the highest in-hospital mortality (32%)

Consistent differences by phenotype were seen. In the derivation cohort, cumulative 28-day mortality was 5% in the α phenotype, 13% in the β phenotype, 24% in the γ phenotype, and 40% in the δ phenotype.

"Across all cohorts and trials, 28-day and 365-day mortality were highest among the δ phenotype vs the other 3 phenotypes (P<0.001)," the researchers wrote. "In simulation models, the proportion of RCTs [randomized controlled trials] reporting benefit, harm, or no effect changed considerably (e.g., varying the phenotype frequencies within an RCT of early goal-directed therapy changed the results from >33% chance of benefit to >60% chance of harm)."

The researchers categorized patients by phenotype from three international trials -- , , and -- that all failed to show significant benefits associated with the treatments being studied.

"The largest changes were seen in the ProCESS trial, which found no benefit from early goal-directed therapy compared with usual care," the authors wrote. "In simulations, when the δ phenotype was increased, early goal-directed therapy was harmful in more than half of the trials."

It is possible that reanalyzing completed trials by phenotype could result in outcomes that were not recognizable in the trial cohorts as a whole, Seymour noted.

Study limitations included the fact that only routinely available clinical data in the EHR were used to identify phenotypes and differences in short-term and long-term prognosis were present across phenotypes, "perhaps due to different features of the validation cohorts, such as the definition of sepsis, demographics, or burden of organ dysfunction," the authors stated.

Seymour and colleagues concluded that additional studies are needed "to determine the utility of these phenotypes in clinical care and for informing trial design and interpretation."

Co-author Derek Angus, MD, MPH, of the University of Pittsburgh, said that a key goal is delivering targeted treatments to patients with sepsis in the way that treatment is now delivered for many cancers.

"You wouldn't give all breast cancer patients the same treatment," he said in a press statement. "Some are positive or negative for different biomarkers and respond to different medications. The next step is to do the same for sepsis that we have for cancer -- find therapies that apply to the specific types of sepsis and then design new clinical trials to test them."

Promising but Not Yet Possible

In an William A. Knaus, MD, of the University of Virginia in Charlottesville, and Richard D. Marks, JD, of Patient Command in McLean, Virginia, noted that "Assuming these phenotypes are unique and do not simply reflect variations in organ system failure or severity of illness, they might be the used in combination with additional clinical data combined with other manifestations of sepsis as measured by systems biology and novel gene expression patterns."

They added that such an approach might identify new subsets of patients with sepsis who require different immunotherapeutic interventions.

However, they cautioned that "this promise of combining advanced machine learning approaches with large clinical and biological data sets is not possible today."

The editorialists pointed out that the authors could access only a small amount of clinical data, and that additional data on variables such as past clinical and pathological diagnoses; exact type and severity of individual comorbidities; and recent interventions that might reduce a patient's ability to mount a strong immune defense would have been useful.

"However, these clinical variables are not readily accessible today, and, moreover, it is uncertain whether and how the addition of these variables would have improved the models the authors derived," they stated. "The study by Seymour and colleagues represents the brave new world of attempting to apply patient data, machine learning, and artificial intelligence to better understand complex, serious clinical problems. However, the ultimate answer to the question 'will this approach improve patient outcomes?' remains unknown."

Disclosures

Seymour disclosed relevant relationships with Edwards and Beckman Coulter. Angus disclosed serving as JAMA associate editor, and disclosed relevant relationships with Ferring Pharmaceuticals, Bristol-Myers Squibb, Bayer AG, Beckman Coulter, and ALung Technologies, as well as having patent applications pending for Selepressin and proteomic biomarkers of sepsis in elderly patients.

Seymour, Angus, and several co-authors disclosed support from the NIH.

Knaus disclosed no relevant relationships with industry. Marks disclosed a relevant relationship with the Health Record Banking Alliance.

Primary Source

JAMA

Seymour CW, et al "Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis" JAMA 2019; DOI: 10.1001/jama.2019.5791.

Secondary Source

JAMA

Knaus WA and Marks RD "New Phenotypes for Sepsis -- The Promise and Problem of Applying Machine Learning and Artificial Intelligence in Clinical Research" JAMA 2019; DOI: 10.1001/jama.2019.5794.