PubMed1997
OBJECTIVE: To determine whether C-reactive protein (CRP) can be used as a parameter to identify the time point when antibiotic treatment can safely be discontinued in a defined major subgroup of neonates treated for suspected bacterial infection. PATIENTS: One hundred seventy-six newborns with birth weights of greater than 1500 g and without central lines and mechanical ventilation who had suspected bacterial infection were enrolled in a prospective study. SETTING: Tertiary care neonatal reference center. DESIGN: Serum concentrations of CRP were determined 24 to 48 hours after the first dose of antibiotics. If CRP levels were less than 10 mg/L, infants were considered unlikely to be infected, and the antibiotic treatment was stopped using CRP as the single decision criterion in 84 of 94 newborns (group 1). Infants with CRP levels of 10 mg/L or greater were considered likely to be infected and randomized to two study groups. In 38 of 39 neonates (group 2a), CRP was determined daily, and antibiotic therapy was discontinued as soon as CRP returned to less than 10 mg/L. Forty-three neonates with likely infection (group 2b) were treated for at least 5 days, and relapse rates of bacterial infections were compared between groups 2a and 2b. OUTCOME MEASURES: The primary outcome variable of the study was the number of infectious relapses of the primary infection. This was assessed by the need for a second course of antibiotics within 4 weeks of the first one. The value of CRP for guiding treatment duration was determined by calculating the negative predictive value of CRP with respect to further treatment in study groups 1 and 2a. Treatment durations and relapse incidence in the two groups of neonates with likely infection (groups 2a and 2b) were compared. RESULTS: Within the 4-week follow-up period, one infant in group 1 and no infant in group 2a received a second course of antibiotics for bacterial infection. CRP levels of less than 10 mg/L determined later than 24 hours after beginning the antibiotic treatment thus correctly identified 120 of 121 infants as not needing further antibiotics. This corresponds to a negative predictive value with respect to further treatment of 99% (95% confidence interval, 95.4% to 99.9%). The mean treatment duration was 3.7 (median, 4; range, 3 to 6) days in the CRP-guided group and 5.5 (median, 5; range, 5 to 7) days in the at least 5-day study group. In the latter group, one infant was treated for a potential relapse, and one infant was treated for a likely relapse. The low relapse rates in both treatment groups are a preliminary indication that relapses may not occur more frequently if patients are treated until CRP is negative rather than for a 5-day or longer treatment period. CONCLUSIONS: We conclude that CRP could be a key parameter for individually guiding the duration of antibiotic treatment in a major subgroup of newborns with suspected bacterial infection. This approach would allow considerably shorter courses of antibiotic therapy.