Abstract
IMPORTANCE: C-reactive protein (CRP) is a commonly used biomarker for sepsis. OBJECTIVE: This study aimed to evaluate CRP's diagnostic performance to detect neonatal early-onset sepsis (EOS), while accounting for both gestational and postnatal age. METHODS: We conducted a multicenter retrospective study of neonates born from January 2006 to December 2017 in public hospitals in Hong Kong. Neonates who had CRP tested within 72 h of birth, prior to antibiotics initiation, were included. CRP values were analyzed by gestational age and time after birth. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves, sensitivity, specificity, predictive values, and likelihood ratios (LRs). RESULTS: Among 100 327 tested neonates, 448 developed culture-confirmed EOS, and 34 had meningitis. In uninfected neonates, CRP concentrations rose naturally after 8 h of life, peaking at 24-32 h, with higher values in term neonates at 24-48 h (median 6.2 mg/L, interquartile range 2.9-14.3). Diagnostic performance was better in preterm neonates (e.g., the area under the ROC curve after 4 h of life: 0.88 in <34 weeks, 0.83-0.92 in 34-36 weeks, and 0.73-0.77 in term neonates). Sensitivity within the first 4 h was low across all groups. LRs increased with higher CRP values in preterm neonates. A CRP value of >12.0 mg/L was associated with increased risk of meningitis. INTERPRETATION: CRP is not recommended for EOS screening in term neonates, but it may be considered for assessing the risk of EOS in preterm neonates. Interpretation must account for gestational and postnatal age.