Soluble TREM-1 as a predictive factor of neonatal sepsis: a meta-analysis

The efficacy of soluble triggering receptor expressed on myeloid cell-1 (TREM-1) in detecting sepsis in adults has already been proven. To date, however, consensus in the field of neonatal sepsis is lacking. The purpose of the present systematic review is to accumulate current evidence in this field. We systematically searched Medline (1966-2017), Scopus (2004-2017), Clinicaltrials.gov (2008-2017), EMBASE (1980-2017), Cochrane Central Register of Controlled Trials CENTRAL (1999-2017) and Google Scholar (2004-2017) along with reference lists from included studies. Eight studies were finally included in the present analysis, with a total number of 667 neonates. The estimated sensitivity for the summary point was 0.95 [95% CI (0.81-0.99)] and the specificity was 0.87 [95% CI (0.56-0.97)]. The diagnostic odds ratio was calculated at 132.49 [95% CI (6.85-2560.70)]. Fagan's nomogram demonstrated that the post-test probability increased to 71% and decreased to 2%, when the pre-test probability was set at 25%. However, significant discrepancy was observed in terms of the used cut-offs; therefore, the sensitivity and specificity presented in our meta-analysis should be reviewed with caution, as they may present an overestimation of the actual predictive efficacy of this protein. Current evidence suggests that sTREM-1 may become a useful biomarker for the prediction of neonatal sepsis. However, the small number of studies and the variation of the threshold values limit its implementation in clinical practice. Future large-scale studies are needed to determine the optimal cut-off value that may discriminate normal levels from those suggestive of the presence of neonatal sepsis.