The accuracy of presepsin in diagnosing neonatal late-onset sepsis in critically ill neonates: a prospective study

Abstract Objectives The diagnostic accuracy of presepsin (P-SEP) in the newborn is still under evaluation. Methods In a multicenter study, we studied the accuracy of P-SEP as a diagnostic marker of late-onset sepsis (LOS) in critical newborns with underlying disorders, to define the most accurate cut-off to distinguish infected from uninfected patients. Results Sixty-nine/351 newborns without infections at admission developed LOS. The median P-SEP value at T0 (admission) was 518.0 ng/L (IQR 313.0–789.0), without significant differences related to underlying diseases (p=0.52). In neonates who developed LOS, P-SEP increased at the onset of infection (T1) (median: 816.0 ng/L) and after 24–48 h (median: 901.0 ng/L) compared with their value at admission (median: 560.0 ng/L) (p<0.01 and p=0.03, respectively). The area under the ROC curve at T1 was 0.71 (95 % CI 0.65–0.78) when all cases of sepsis were included in the analysis and increased to 0.74 (95 % CI 0.66–0.81) considering only confirmed sepsis. Approximately two-thirds of patients were correctly classified, setting the cut-off at 713 ng/L, with a negative predictive value of 89.0 %. Conclusions At a cut-off of 713 ng/L, P-SEP has good accuracy in diagnosing LOS in critically ill newborns. In uninfected newborns, the median value of P-SEP is not influenced by any underlying pathology.