Abstract. BACKGROUND: Cytokines might be helpful to diagnose late onset sepsis (LOS) in newborn infants. Many studies on cytokines did not discriminate culture-proven from clinically-suspected sepsis; however, such differentiation is clinically useful. OBJECTIVES: To evaluate the feasibility to differentiate among culture-proven LOS, clinical LOS and controls using a battery of cytokines. STUDY DESIGN: This prospective study was conducted at the NICU of Harapan-Kita Women and Children's Hospital, Jakarta-Indonesia. Three groups of infants with postnatal age >72 hours of age were enrolled in the study: culture-proven sepsis group (PS) (n = 18), clinical sepsis group (CS) (n = 25) and control group (n = 34). A battery of 25 cytokines was measured in each infant five times: at enrollment, after 4 hrs, 12 hrs, 24 hrs, and 48 hrs using Invitrogen-immunoassays-Luminex™100. RESULTS: There were no significant differences in gestational age or mode of delivery among the three groups. IL-1β, IL-2r, IL-6, IL-8, IL-10 and MIP-1a were significantly higher at all measurement points in group PS compared to controls. IL-13 was lower at all measurement moments in group CS compared to controls, IL-12 was lower and IP-10 higher between 0 and 24 hrs. IL-1Ra, IL-6, IL-8, IL-13, IL-15, TNFα, MIP-1a and MIP-1b were higher at all the measurement moments in group PS compared to group CS. The ROC curves show that IL-6, IL-8, IL-15, MIP-1a, MIP-1b and TNFα have a sensitivity and specificity between 80 and 85% during the first 24–48 hours after the onset of infection. IL-6, IL-15, MIP-1a, MIP-1b and TNFα showed the best likelihood ratios. CONCLUSIONS: IL-6, IL8, IL 15, MIP-1a, MIP-1b and TNFα are potentially good markers for detecting a proven LOS. In case these cytokines are not elevated in sick infants, other causes than an infection have to be identified.