Abstract
Background Early onset sepsis (EOS) remains a significant cause of neonatal mortality and morbidity. About 10% of term neonates were exposed to antibiotics for EOS risk, despite a very low incidence (around 1‰) in this group especially after the introduction of intrapartum antibiotic prophylaxis. One of the main reasons for over treating the neonates is the lack of ideal diagnostic biomarkers. C-reactive protein (CRP) and complete blood count (CBC) are most commonly used biomarkers for EOS, but the diagnostic performance of CRP was poorly understood with most studies on late onset sepsis. In the past 15 years, our neonatal unit has used CRP systematically in neonates with risk or signs of EOS, providing a good opportunity to analyze its diagnostic accuracy.
Objectives In this study, we aimed to evaluate the performance of CRP for early onset infection diagnosis in symptomatic neonates and screening in asymptomatic high-risk neonates.
Methods This is a retrospective study done in a tertiary neonatal care center in Hong Kong. Term neonates born during the period of 01/01/2005–30/06/2018 with blood taken for CRP testing in the first 72 hours were included. Their CRP results were included into the analysis if blood were taken before antibiotic treatment. Subjects were divided into four subgroups according to their infectious status and having clinical signs or not. CRP values were compared between neonates infected and not infected. Receiver operating characteristic (ROC) curves were constructed separately in symptomatic and asymptomatic groups. The sensitivity, specificity, positive predict value, negative predict value, likelihood ratios and post-test probability were determined using different cut-off values.
Results 24,344 infants with 28,830 CRP results were included. Early onset infection was confirmed in 75 (0.31%) cases, 68 of whom had EOS and 7 had local infections (urinary tract or eyes). Discrimination improved after 8 hours of birth in asymptomatic neonates (the area under the ROC curve (AUC) was 0.49 at <8 hours and 0.79 at ≥8 hours), and 24 hours of birth in symptomatic neonates (the AUC was 0.65 at <24 hours and 0.75 at ≥24 hours). High CRP values were less informative in symptomatic neonates than asymptomatic neonates (eg. the likelihood ratio for CRP>10 mg/L was 2.2 in symptomatic and 5 in asymptomatic neonates). Using cut off value of 10 mg/L at ≥24 hours, sensitivity was 72% in symptomatic neonates and 83% in asymptomatic neonates, and specificity was 69% and 86% respectively. Positive predict values were very low (0.3–6%).
Conclusions Diagnostic performance of CRP was poor within 24 hours of birth. However, in asymptomatic neonates, high CRP in 8–24 hours was informative of infection. Because of the low positive predict value, attention should be paid to overtreatment when using CRP for decision making.
