Novel blood protein biomarkers for distinguishing bacterial from non-bacterial infection in children: a case-control study
Drummond, H.; Roarty, C.; Groves, H.; Waterfield, T.; Mills, C.
Show abstract
BackgroundDistinguishing bacterial from viral or inflammatory conditions in hospitalised children is clinically challenging due to overlapping clinical features and limitations of existing diagnostics. Culture-based tests are slow, and commonly used biomarkers such as C-reactive protein (CRP) lack sufficient accuracy, contributing to empirical antibiotic use and antimicrobial resistance. We aimed to evaluate the diagnostic performance of novel blood protein biomarkers for bacterial infection in children admitted with infection or inflammatory illness. MethodsWe performed a prospective case-control study of children [≤]16 years admitted to the Royal Belfast Hospital for Sick Children (2020-2023) with bacterial infection, viral infection or inflammatory illness. Plasma samples were analysed for previously reported novel biomarkers and signatures (LCN2; TRAIL+IP-10+CRP; E-selectin+IL18+ NCAM1+LCN2+IFN-{gamma}+LG3BP and E-selectin+IL18+ NCAM1+LCN2+IFN-{gamma}). The diagnostic accuracy of individual biomarkers and biomarker signatures was assessed using ROC curves. ResultsFifty-two children were included (13 bacterial, 20 viral, 19 inflammatory). All evaluated biomarker signatures and LCN2 distinguished bacterial from viral infection (AUC 0.819- 0.935), with the TRAIL+IP-10+CRP signature achieving the highest accuracy (AUC 0.935). In bacterial-inflammatory comparisons, performance was lower (AUCs 0.607-0.745); the E-selectin+IL-18+NCAM1+LCN2+IFN-{gamma} signature performed best (AUC 0.745) and outperformed CRP (AUC 0.595). A novel three-protein signature (E-selectin+LCN2+IFN-{gamma}) had a significantly higher AUC than CRP for distinguishing bacterial infections from non-bacterial (viral and inflammatory combined). ConclusionsSeveral novel host protein biomarkers and signatures had a higher diagnostic accuracy than CRP for differentiating bacterial from non-bacterial illness (viral and inflammatory) in hospitalised children. These findings support the potential of biomarker-guided diagnostics to improve accuracy and facilitate earlier antibiotic de-escalation.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Machine Learning Identifies Complicated Sepsis Trajectory and Subsequent Mortality Based on 20 Genes in Peripheral Blood Immune Cells at 24 Hours post ICU admission 94%
- Loss of Factor H family proteins associates with meningococcal disease severity 94%
- Immune-Based Prediction of COVID-19 Severity and Chronicity Decoded Using Machine Learning 93%
Similar papers in this journal
- Development and validation of a cellular host response test as an early diagnostic for sepsis 94%
- Clustering ICU patients with sepsis based on the patterns of their circulating biomarkers: a secondary analysis of the CAPTAIN prospective multicenter cohort study 94%
- Induction of an early IFN-γ cellular response and high plasma levels of SDF-1α are inversely associated with COVID-19 severity and residence in rural areas in Kenyan patients 94%
Similar papers in this journal
Similar papers in this journal
- Plasma protein biomarkers distinguish Multisystem Inflammatory Syndrome in Children (MIS-C) from other pediatric infectious and inflammatory diseases 97%
- A tool to distinguish viral from bacterial pneumonia 93%
- COVID-19 and Multisystem Inflammatory Syndrome in Latin American children: a multinational study 92%
Similar papers in this journal
- Standard blood laboratory results in SARS-CoV-19 positive patients: do they show a typical pattern? 94%
- COMPARISON OF sPLA2-IIA PERFORMANCE WITH HIGH-SENSITIVE CRP, NEUTROPHIL PERCENTAGE, PCT AND LACTATE TO IDENTIFY BACTERIAL INFECTION: A PROSPECTIVE STUDY 93%
- Calprotectin as a sepsis diagnostic marker in critical care: a retrospective observational study 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.