Background It is common clinical practice to repeat serum C reactive protein (CRP) estimation in the first 48 h after starting empirical antibiotics for neonatal sepsis. The change in CRP is believed to indicate whether the empirical antibiotics are appropriate or not, but there is little evidence to support this practice.

Methods This was a nested case–control study on neonates with suspected sepsis (clinical signs+baseline CRP >10 mg/L). We drew samples for serum CRP at baseline and at 24, 36 and 48 h after starting empirical antibiotics and stored them at −20°C. Those with positive blood cultures were enrolled into two groups: those who had received empirical antibiotics to which the organism was sensitive (‘sensitive’) and those who had received antibiotics to which the organism was resistant (‘resistant’). Stored samples of these subjects were assayed for CRP. Repeated CRP values were compared between groups by mixed linear models. We evaluated change in CRP from baseline as a diagnostic test for identifying empirical use of sensitive antibiotics.

Results We enrolled 45 and 44 subjects in ‘sensitive’ and ‘resistant’ groups, respectively. In the ‘resistant’ group, median CRP increased with time but decreased in the ‘sensitive’ group. Decline in CRP by 48 h identified the use of antibiotics to which the organism was sensitive with 89% sensitivity and 80% specificity.

Conclusions Serial CRP values in the first 48 h of antibiotic therapy help to predict whether the causative organism is sensitive to the empirical antibiotics administered.