Introduction

Morbidity and mortality of nosocomial/late-onset sepsis (LOS) still remain high, especially in preterm neonates [1, 2]. Accurate and timely diagnosis of sepsis is clinically difficult in the neonatal period. In addition, the reliability of existing diagnostic tests for early identification of septic neonates varies, mainly when used separately [3].

Triggering receptor expressed on myeloid cells (TREM)-1 is a 30-kDa transmembrane glycoprotein expressed selectively on blood neutrophils, macrophages, and mature monocytes; it plays an important role in the amplification of inflammation. TREM-1 triggers the release of proinflammatory cytokines, increases surface expression of cell receptors, and activates neutrophil degranulation and oxidative burst [4]. During endotoxemia, TREM-1 expression on effector cells is markedly increased in human blood, as is its soluble form sTREM-1 [5]. In this context, sTREM-1 has been evaluated in blood of adults with sepsis [6] and community-acquired pneumonia [7] as well as in bronchoalveolar lavage and cerebrospinal fluid of patients with ventilator-associated pneumonia [8] and meningitis [9], respectively, and it was also found useful in distinguishing infection. However, there are currently no available data concerning septic neonates.

The aim of this study was to determine serum sTREM-1 levels in neonates and assess their possible value in early diagnosis of LOS against interleukin-6 (IL-6), another commonly used sepsis biomarker with high diagnostic value in detecting bacterial infection in neonates [10, 11].

Methods

This prospective observational study was performed in a single, level III neonatal intensive care unit after approval by the human research review board of our institution. The study population included preterm and term neonates evaluated for nosocomial/LOS (>day 3 of life). Informed consent was obtained from the parents before enrollment. Neonates born to mothers with clinical chorioamnionitis or those who had early-onset sepsis, congenital infections, or anomalies were excluded. Eligible neonates were enrolled once clinical deterioration suggestive of sepsis occurred. Initial laboratory investigations included white blood cell count and differential, platelet counts, C-reactive protein (CRP), arterial blood gas, as well as blood and urine cultures. Neonates were later classified into the groups of infected and noninfected ones. The former included neonates with confirmed (positive blood cultures for microbes or fungi) and possible sepsis (clinical and laboratory evidence of sepsis but negative blood cultures). Laboratory evidence of infection was metabolic acidosis, thrombocytopenia, leukopenia/leukocytosis, immature/total neutrophil ratio >0.2, and CRP >10 mg/L. Neonates with negative blood cultures and no laboratory evidence of infection were classified as noninfected (nonseptic). Data recorded included gestational age (GA), birth weight (BW), gender, survival, and microorganisms isolated from blood cultures. Blood was drawn at time of initial laboratory evaluation for sepsis and centrifuged, and sera were stored at -70°C until analysis. Serum sTREM-1 and IL-6 were measured by commercially available enzyme-linked immunosorbent assays (Human TREM-1 Quantikine ELISA Kit, NEW and Human IL-6 Quantikine HS ELISA Kit; R&D Systems, Minneapolis, USA).

Numerical data are expressed as median with range. Mann–Whitney U, Kruskal–Wallis, and Fisher’s exact tests were used for comparisons, as appropriate. Multiple regression analysis was used to explore the relationship between a continuous dependent variable and a set of independent variables. Receiver operating characteristic (ROC) curves were constructed considering infected versus noninfected neonates. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and likelihood ratio (LR) were calculated based on cutoff values derived from ROC analysis. Values <0.05 were considered statistically significant. Statistical analysis was performed using SPSS 15.0 (SPSS Inc., Chicago, IL) for Windows and STATA 9.2 (StataCorp LP, College Station, TX).

Results

Our study population comprised 52 neonates: 21 noninfected and 31 infected (9 with possible and 22 with confirmed LOS). Of the enrolled neonates, 37 (71.1%) were preterm and 15 (28.9%) were term. Noninfected neonates had significantly lower GA [30 (24–39) weeks versus 35 (24–40) weeks, p = 0.035] and BW [1,190 (720–4,785) g versus 1,990 (890–3,600) g, p = 0.025] as compared with infected neonates. Male sex was equally distributed between groups (57.1% versus 61.2%, in noninfected and infected group, respectively; p = 0.78). Most common microorganisms isolated from blood cultures were Gram-positive (n = 13) followed by Gram-negative microbes (n = 7) and fungi (n = 2). All urine cultures were negative. Mortality was significantly lower in noninfected neonates (4.8% versus 22.5%, p = 0.033).

Infected neonates had significantly higher serum sTREM-1 and IL-6 as compared with the noninfected ones (Fig. 1). Neonates with either confirmed or proven sepsis had significantly higher sTREM-1 and IL-6 when compared with noninfected neonates. However, no difference in sTREM-1 and IL-6 levels was found between neonates with possible and confirmed sepsis (Fig. 1). Multiple regression analysis showed that infection was a significant independent predictor for both sTREM-1 (p = 0.033) and IL-6 levels (p < 0.001) after adjusting for GA and BW. Also, sTREM-1 and IL-6 were comparable between infected preterm and term neonates, and between noninfected preterm and term ones (data not shown). ROC curve analysis revealed that both sTREM-1 and IL-6 resulted in significant areas under the curve (AUC) with respect to early identification of infected neonates. Although this was larger for IL-6 [AUC = 0.892, 95% confidence interval (CI) (0.808–0.976), p = 0.001] than for sTREM-1 [AUC = 0.733, 95% CI (0.587–0.880), p = 0.005], the difference between the two AUC was not significant (p = 0.053). Optimal cutoff level, sensitivity, specificity, PPV, NPV, and LR of serum sTREM-1 and IL-6 as well as of their combination in predicting sepsis are shown in Table 1.

Fig. 1
figure 1

Serum sTREM-1 and IL-6 levels (median with range) in noninfected and infected neonates (a). Values of neonates with possible and confirmed sepsis, forming the group of infected neonates, are shown separately, as compared with noninfected ones (b). a p < 0.05 possible sepsis versus no sepsis, b p < 0.05 confirmed sepsis versus no sepsis

Full size image
Table 1 Diagnostic accuracy of serum sTREM-1 and IL-6 levels as well as of their combination
Full size table

Discussion

This study showed that sTREM-1 and IL-6 increase significantly in the serum of infected preterm and term neonates with possible and confirmed LOS, and that sTREM-1 has a significant diagnostic performance which, however, is not better than that of IL-6 alone.

Studies in critically ill adults demonstrated significantly increased plasma sTREM-1 in patients with sepsis as compared with those with noninfectious systemic inflammatory response syndrome. In addition, sTREM-1 was reported to be better than procalcitonin and CRP as biomarker in detecting sepsis [6]. Nevertheless, recent investigations proved increased circulating sTREM-1 levels in acute inflammation without infection [12]. Our data, which to the best of our knowledge are the only to report on sTREM-1 in neonates, support the elevation of the specific protein in serum upon infection.

Results of this study confirm the high diagnostic value of IL-6 in neonates, mainly when measured early during the course of LOS [10]. In addition, our data provide evidence that sTREM-1 may be used for assessing neonatal nosocomial infections, although it was not found to be more accurate than IL-6 in differentiating septic neonates. This fact is in line with a recent meta-analysis involving 13 studies in adults, where sTREM-1 was found to be reliable in bacterial infection with sensitivity of 0.82 and specificity of 0.86 [13]. Similarly, plasma sTREM-1 was documented as having better sensitivity and specificity than CRP in predicting bloodstream infections in febrile infants [14]. In our study, ROC curve analysis using CRP was not performed because it was taken into consideration for classification of neonates as noninfected/infected. Interestingly, despite the elevation of both serum sTREM-1 and IL-6, and contrary to others [15], we failed to document any relationship between them. Regarding diagnostic accuracy, the combination of sTREM-1 plus IL-6 increased both their sensitivity and NPV, which is desirable in conditions such as neonatal sepsis [10]. As, however, LR and the more clinically useful PPV [16] decreased compared with IL-6 alone, it can be concluded that diagnostic performance is virtually not improved when the specific sepsis markers are combined. It should be noted, though, that the predictive value of a test is determined by the prevalence of the disease, and thus our results may not be applicable to populations of neonates with different prevalence of LOS [16].

A limitation of the study was that neonates with possible sepsis were assumed to be infected similarly to neonates who had culture-proven sepsis. However, this group unification is justified as sTREM-1 levels were comparably high in these groups and, furthermore, IL-6 increased markedly in seven out of the nine neonates with possible sepsis, by far above the proposed cutoff levels of 31–70 pg/mL for predicting neonatal sepsis [10, 17]. In addition, neonatal sepsis cannot be ruled out solely on the basis of a negative blood culture result [18].

In conclusion, serum sTREM-1 levels increase in infected neonates. However, diagnostic accuracy of a single measurement of circulating sTREM-1 is not better than that of IL-6 alone. Moreover, the combination of the two markers does not improve their accuracy in detecting infected neonates. The importance of sTREM-1 measurement in septic patients including neonates should be evaluated in the context of large multicenter trials, especially after exploring multiple cytokine interactions.