A blood test may successfully predict death from sepsis, according to a new study published in the July 24 issue ofScience Translational Medicine.

The study, led by Raymond J. Langley, PhD, from the National Center for Genome Resources, Santa Fe, New Mexico, evaluated 292 patients admitted to emergency departments between 2005 and 2009 with suspected community-acquired sepsis, defined as acute infection and 2 or more systemic inflammatory response syndrome criteria. This group included patients from 3 tertiary care hospitals: an initial discovery cohort and 2 validation patient sets. They collected data regarding physical examination, medical history, and acute illness scores (Acute Physiology and Chronic Health Evaluation II [APACHE II] and Sequential Organ Failure Assessment [SOFA]) at enrollment (t 0) and 24 hours later (t 24). They also measured mass-to-charge ratios of 439 plasma metabolites from blood samples taken at t 0 and t 24, and then compared metabolite z-scores between survivors and nonsurvivors.

The researchers used the metabolomics data, along with clinical features, age, mean arterial pressure, hematocrit, and temperature, to develop an algorithm that predicted patient survival.
The biomarker panel consisted primarily of carnitine esters with medium- or short-chain fatty acids and branched-chain amino acids. The researchers noted that plasma concentrations of the carnitine esters were decreased in sepsis survivors and elevated in sepsis nonsurvivors relative to controls. Fatty acid levels were also elevated among the sepsis nonsurvivors.

"A logistic regression model using carnitine esters and clinical variables consistently categorized survivors with greater than 85% accuracy, whereas sepsis nonsurvivors were accurately predicted with 45 to 55% accuracy in most of the test sets," Dr. Langley and colleagues write.

The researchers also found that there were no major differences between patients with infections withStreptococcus pneumonia, Escherichia coli, or Staphylococcus aureus.

The authors acknowledge that because testing was performed at the time of presentation at an emergency department, the prognostic value of this panel may be limited. In addition, because the differences between survivors and nonsurvivors increased over time, serial testing in individual patients may be more beneficial in differentiating those patients with poor outcomes. The authors also note that testing at more than 2 times and in a larger population is needed to further assess the utility of this biomarker panel.

When asked in an interview with Medscape Medical News whether this metabolomics model would be useful in the clinical setting, Loren Miller, MD, MPH, director of the Infection Prevention and Control Program, Harbor–University of California, Los Angeles, Medical Center in Torrance, said: "It could possibly become a tool in the future, but there are many limitations and barriers before it would have practical value."

"Predicting death or survival from sepsis is of modest clinical utility," said Dr. Miller. "It may have a role in research to describe populations or confounders that correlate with clinical outcomes to perhaps predict if an intervention (ie, a new antibiotic) to improve survival helped patients because you would know what their prognosis was."

Funding for this study was provided by grants from the National Institutes of Health, Pfizer Inc, and Roche Diagnostics Inc. Two of the authors are consultants for bioMérieux. One author has received honoraria in the last 3 years from Achaogen, Arpida, Astellas Pharma Inc, Cubist Pharmaceuticals, Durata, Inhibitex, Leo Pharma, Merck & Co Inc, Pfizer, Targanta, Theravance Inc, and Ortho-McNeil and is a paid consultant for Affinium, Astellas Pharma Inc, Biosynexus, Novartis, Cubist Pharmaceuticals, Inimex, Merck & Co Inc, Galderma, Johnson & Johnson, Medicines Company, and NovaDigm. Several authors have received a patent related to this work and have further patent applications pending. Dr. Miller has disclosed no relevant financial relationships.

Sci Transl Med. 2013;5:195ra95. Abstract