Diagnostic accuracy of HMGB-1, sTREM-1, and CD64 as markers of sepsis in patients recently admitted to the emergency department.

OBJECTIVES: The objectives were to evaluate the diagnostic accuracy for sepsis in an emergency department (ED) population of the cluster of differentiation-64 (CD64) glycoprotein expression on the surface of neutrophils (nCD64), serum levels of soluble triggering receptor expressed on myeloid cells-1 (s-TREM-1), and high-mobility group box-1 protein (HMGB-1).

METHODS: Patients with any of the following as admission diagnosis were enrolled: 1) suspected infection, 2) fever, 3) delirium, or 4) acute hypotension of unexplained origin within 24 hours of ED presentation. Levels of nCD64, HMGB-1, and s-TREM-1 were measured within the first 24 hours of the first ED evaluation. Baseline clinical data, Sepsis-related Organ Failure Assessment (SOFA) score, Acute Physiology and Chronic Health Evaluation (APACHE II) score, daily clinical and microbiologic information, and 28-day mortality rate were collected. Because there is not a definitive criterion standard for sepsis, the authors used expert consensus based on clinical, microbiologic, laboratory, and radiologic data collected for each patient during the first 7 days of hospitalization. This expert consensus defined the primary outcome of sepsis, and the primary data analysis was based in the comparison of sepsis versus nonsepsis patients. The cut points to define sensitivity and specificity values, as well as positive and negative likelihood ratios (LRs) for the markers related to sepsis diagnosis, were determined using receiver operative characteristics (ROC) curves. The patients in this study were a prespecified nested subsample population of a larger study.

RESULTS: Of 631 patients included in the study, 66% (95% confidence interval [CI] = 62% to 67%, n = 416) had sepsis according with the expert consensus diagnosis. Among these sepsis patients, SOFA score defined 67% (95% CI = 62% to 71%, n = 277) in severe sepsis and 1% (95% CI = 0.3% to 3%, n = 6) in septic shock. The sensitivities for sepsis diagnosis were CD64, 65.8% (95% CI = 61.1% to 70.3%); HMGB-1, 57.5% (95% CI = 52.7% to 62.3%); and s-TREM-1, 60% (95% CI = 55.2% to 64.7%). The specificities were CD64, 64.6% (95% CI = 57.8% to 70.8%), HMGB-1, 57.8% (95% CI = 51.1% to 64.3%), and s-TREM-1, 59.2% (95% CI = 52.5% to 65.6%). The positive LR (LR+) for CD64 was 1.85 (95% CI = 1.52 to 2.26) and the negative LR (LR-) was 0.52 (95% CI = 0.44 to 0.62]; for HMGB-1 the LR+ was 1.36 (95% CI = 1.14 to 1.63) and LR- was 0.73 (95% CI = 0.62 to 0.86); and for s-TREM-1 the LR+ was 1.47 (95% CI = 1.22 to 1.76) and the LR- was 0.67 (95% CI = 0.57 to 0.79).

CONCLUSIONS: In this cohort of patients suspected of having any infection in the ED, the accuracy of nCD64, s-TREM-1, and HMGB-1 was not significantly sensitive or specific for diagnosis of sepsis.