Abstract Background : Despite the high prevalence of late-onset sepsis (LOS) in neonatal intensive care units (NICUs), a reliable diagnosis remains difficult. The time needed to obtain laboratory results of biomarkers such as C-reactive protein (CRP) and procalcitonin (PCT) and blood culture explains why an unjustified antibiotic use is observed in numerous hospitalized neonates. This results in an increased frequency of antibiotic resistance, microbiota modification, and neonatal complications. The objective of EMERAUDE study was to identify biomarkers (alone or in combination) to early exclude the diagnosis of LOS in neonates with suggestive clinical signs. Methods : A prospective, multicenter cohort study (EMERAUDE)was conducted in 2 French NICUs. The participants were hospitalized neonates at ≥7 days of life with signs of suspected LOS enrolled from November 2017 to November 2020. Serum samples were collected during the venipuncture prescribed for blood culture. Eleven biomarkers were measured using customized multiplexed assays in the ELLA Automated Immunoassay System (ProteinSimple, San Jose, CA, USA) for PCT, IP-10, IL-6, IL-10, NGAL, PTX3, presepsin and LBP, and using conventional ELISA for calprotectin (R&D Systems, Minneapolis, MN, USA), gelsolin(Elabsciences, Houston, TX, USA) and IL-27(R&D Systems, Minneapolis, MN, USA). An independent adjudication committee, blind to biomarkers, assigned each patient to either infected, not infected or unclassified groups. Performances of biomarkers were assessed considering a sensitivity of at least 0.898. Results : A total of 230 patients were analyzed. They were mainly preterm (80%) with a median gestational age of 27 weeks and a median birth weight of 940 grams. The adjudication committee classified 22% of patients (51/230) as infected and all of these received antibiotics. Among patients of the not infected group, 27% (42/153) also received antibiotics. The best biomarkers alone were IL-6, IL-10 and NGAL; the area under the curve [95%CI] was, respectively, 0.864 [0.798-0.929], 0.845 [0.777-0.914], and 0.829 [0.760-0.898]. Combinations of up to 4 biomarkers were analyzed and the best were PCT/IL-10, PTX3/NGAL, and PTX3/NGAL/gelsolin. The best models of biomarkers could avoid up to 64% of unjustified antibiotics. Conclusions : At the onset of clinical suspicion of LOS, the dosing of additional biomarkers could help the clinician in identifying not infected patients. Trial registration : ClinicalTrials.gov ID: NCT03299751. Registered 3 October 2017.