Can a negative procalcitonin level guide antibiotic therapy in early-onset neonatal sepsis?

Background. Procalcitonin (PCT) has been used in the diagnosis of early-onset neonatal sepsis (EONS) in conjunction with other markers of infection, and levels are highest at the onset of infection and decline over time. This study evaluated whether an initial negative PCT level could be used to withhold antibiotics in neonates presenting with suspected EONS and whether PCT levels differed between premature and term babies. Methods. Neonates undergoing evaluation for suspected sepsis in the Neonatal Unit, Johannesburg Hospital, within 24 hours of birth between July and September 2004 were included. Patients were categorised into various categories of infection using risk factors for infection, white cell count, platelet count, C-reactive protein and blood culture results. Babies were started on empiric parenteral antibiotics as per unit protocols. PCT was correlated with infection categories. Results. The final analysis included 194 babies, 131 premature and 63 term; 145 had 'no infection', 47 'probable infection' and 2 'definite infection'. The mean PCT levels (and ranges) for the three categories were 1.6 ng/ml (range 0.5 - 37.5 ng/ml), 11.9 ng/ml (0.5 - 150.4) and 6.7 ng/ml (0.5 - 12.9), respectively. Using a cut-off of 0.5 ng/ml, the negative predictive value (NPV) of PCT was 80% and the positive predictive value (PPV) 39%. Increasing the cut-off of PCT had no effect on the NPV. Receiver operating characteristic (ROC) analysis had an area under the curve of 0.631. Conclusions. The NPV of PCT on admission for suspected EONS is better than the PPV, but not sufficiently reliable to exclude sepsis, even when using higher cut-off values. PCT levels did not differ significantly between premature and term babies.