Background: Neonatal studies reporting the diagnostic performance of C-reactive protein (CRP) often use “clinical sepsis” as the outcome of interest, despite the subjective nature of this diagnosis. Blood or CSF culture is the diagnostic standard for neonatal early-onset sepsis (EOS). As the incidence of culture-confirmed EOS has declined, the statistical performance of CRP in predicting EOS and the clinical impact of its use in EOS management is questioned. While some studies report use of serial CRP to allow timely discontinuation of antibiotics, others report increases in antibiotic initiation and duration when determining antibiotic therapy on CRP results. Objective: To determine the performance of CRP in the diagnosis of culture-confirmed EOS; and to determine the difference in antibiotic use among infants without EOS in relation to CRP measurements. Methods: Retrospective cohort study of infants born at two perinatal sites from 2009-2014 with a blood culture (BC) obtained at ≤72 hours of age. For performance diagnostics, we included the first BC obtained for an infant, and paired CRP level obtained within ±4 hours of the BC. Antibiotics initiated ≤72 hours of age were included and prolonged antibiotics were defined as duration >48 hours. EOS was defined as blood or CSF culture obtained ≤72 hours of age and growing a pathogenic species. Abnormal CRP value was defined, per center policy, as ≥10 mg/L. All analysis was performed using STATA. Results: Of 53,262 infants born during study period, 7,552 (14.2%) had a BC obtained at ≤72 hours of age and 41 infants were diagnosed with EOS (0.8 cases per 1000 live births). Study cohort comprised of 4,268/7,552 (56.5%) infants with a paired CRP value (Table 1) and 24 cases (0.6%) of EOS. Performance of CRP compared to paired BC is shown in Table 2. Per site practice, majority of infants (4,084, 95.7%) with a BC were also initiated on antibiotics. Among infants with a negative BC, infants with CRP ≥10 mg/L were significantly more likely to receive prolonged antibiotics (252/434, 58.1%) compared to infants with CRP <10 mg/L (1,282/3,810, 33.7%), p <0.001. Conclusion: EOS is a low incidence disease and the modest positive likelihood ratio of CRP at a cut-off ≥10 mg/L does not improve the post-test probability of a true diagnosis. Despite this, infants with elevated CRP levels were significantly more likely to receive prolonged antibiotic courses in the absence of culture-confirmed EOS. The use of CRP in neonatal early onset sepsis management should be questioned.

Figure II

Overall Mortality Rate Forest Plot

All thirteen included studies reported data on all-cause mortality. Five hundred thirty participants were included in the analysis. There were 41 deaths from the 282 participants in the treatment group while there were 54 deaths from the 248 participants in the control group. There was a significant decrease in mortality in the treatment group (RR 0.69, 95% CI 0.48 to 0.99, Fixed, P = 0.32, I2 = 13%).

Figure II

Overall Mortality Rate Forest Plot

All thirteen included studies reported data on all-cause mortality. Five hundred thirty participants were included in the analysis. There were 41 deaths from the 282 participants in the treatment group while there were 54 deaths from the 248 participants in the control group. There was a significant decrease in mortality in the treatment group (RR 0.69, 95% CI 0.48 to 0.99, Fixed, P = 0.32, I2 = 13%).

Close modal

* Negative Blood Culture cohort includes 9 cases of contaminants paired with a CRP value <10 mg/L; † 24 cases of EOS includes 7 cases of E. coli, 12 cases of Group B Streptococcus and 1 case each of C. koseri, H. influenzae, S. pneumonae, M. morganii and Group D streptococcus.

* Negative Blood Culture cohort includes 9 cases of contaminants paired with a CRP value <10 mg/L; † 24 cases of EOS includes 7 cases of E. coli, 12 cases of Group B Streptococcus and 1 case each of C. koseri, H. influenzae, S. pneumonae, M. morganii and Group D streptococcus.

Close modal