Abstract Background Time-sensitive identification of sepsis plays an essential role in minimizing sepsis-related death. One of the latest and important markers recognized is presepsin, also known as sCD14-ST. After the publication of the Sepsis-3 criteria in 2016, there have been discrepancies found among meta-analyses using both old and new criteria combined. The purpose of this current analysis is to evaluate the performance of presepsin at presentation, using a scientific approach and only considering studies conducted using Sepsis-3 criteria. Methods Electronic searches of PubMed, Scopus, Embase, Web of Science, and the Cochrane Library were carried out until 10 December 2025. Prospective and retrospective studies that expressed the sensitivity and specificity of plasma presepsin for sepsis diagnosis in adults (aged ≥ 18 years) based on Sepsis-3 criteria were considered for review. Bivariate random-effects model estimates of sensitivity, specificity, DOR, and SROC plot were calculated. Results Eleven studies (2248 participants) were included. Pooled sensitivity was 0.782 (95% CI: 0.755–0.807), and pooled specificity was 0.739 (95% CI: 0.714–0.763). The area under the SROC curve was 0.85 (good diagnostic accuracy). The negative likelihood ratio was 0.297 (95% CI: 0.240–0.367), indicating a useful rule out test. Heterogeneity was substantial (I² > 75% for sensitivity and specificity). There was no evidence of a threshold effect based on Spearman correlation ( p = 0.326), although Moses’ model suggested a mild threshold effect ( p = 0.018). Meta regression showed that the cut off value did not significantly influence diagnostic accuracy ( p = 0.878). Conclusions Presepsin has a good diagnostic accuracy for sepsis in adults and is a useful rule-out test. Presepsin has good diagnostic accuracy for sepsis in adults and is a useful rule out test. However, wide variation in cut off values across studies underscores the need for standardization before widespread clinical adoption.
Damavandi et al. (Fri,) studied this question.