Although advances in neonatal intensive care have led to improved survival of very low birth weight (VLBW) and extremely premature infants, late onset sepsis (systemic infection after 48 hours of age) continues to be a significant cause of morbidity and mortality. The incidence of late onset sepsis increases with both decreasing birthweight and gestational age, and has been reported as occurring in approximately 25% of VLBW infants (Stoll 2002; Rubin 2002; Isaacs 1996). Infants with the lowest birth weights are also more likely to have multiple episodes of sepsis (Stoll 2002). In developing countries infection is estimated to cause 30 ‐ 40% of neonatal deaths (WHO 1999). The spectrum of organisms responsible for early onset (vertically transmitted) sepsis differs from that associated with late onset (nosocomial) sepsis. This pattern becomes apparent from day two onwards (Isaacs 1996). Nosocomial infections are frequently associated with clinical deterioration including increased apnoea or ventilatory requirement, temperature instability, abdominal distension, acidosis, lethargy, septic shock, necrotising enterocolitis, meningitis and death (Craft 1999). The complications of necrotising enterocolitis and meningitis predispose an infant to an increased risk of future neurological impairment (Blair 1982; Waugh 1996) and the mortality from late onset sepsis remains high at 7 ‐ 10% (Stoll 2002; Isaacs 1996). These infections are often particularly distressing for infants, parents and staff as they affect VLBW infants who have survived early causes of mortality but remain at risk for ongoing infection. This risk is secondary to: immature immune responses, poorly developed skin and mucosal barriers to infection, numerous entry portals for organisms via cannulae, catheters and endotracheal tubes and continuing exposure to oppurtunistic organisms during often a prolonged hospital stay.

The range of organisms causing late onset sepsis includes gram positive and gram negative bacteria as well as fungal infection. As bacterial infections predominate, empiric antibiotic regimens focus on cover for both gram positive and negative bacterial infection. These antibiotics can be either narrow or broad spectrum in the range of organisms that they target. The epidemiology of late onset infection differs between developing and developed countries in the incidence of infection, the organisms responsible, and the subsequent mortality rates. Historical reviews have also demonstrated that the predominant organisms responsible for neonatal sepsis have changed with time (Stoll 1996).

In developing countries the most common gram positive organisms isolated from neonatal blood cultures taken from babies under 90 days of age are streptococcus pneumoniae, staphlyococcus aureus and streptococcus pyogenes. The most frequent gram negative organisms isolated are E coli and salmonella spp (WHO 1999). WHO recommend that initial treatment of suspected neonatal sepsis should be with penicillin and gentamicin. This regimen covers most of the likely organisms but has poor coverage of both salmonella and the increasingly penicillin resistant staphylococcus aureus.

In developed countries staphylococcus aureus was previously responsible for the majority of late onset infections in many neonatal units with other commonly isolated organisms being coagulase negative staphylococci, E coli, group B streptococcus, Klebsiella pneumoniae, enterococcus, candida and pseudomonas. Coagulase negative staphylococci(CoNS) have now emerged as the leading cause of late onset sepsis in almost all industrialised countries and account for > 50% of positive blood cultures (Rubin 2002; Isaacs 1996). As skin commensals, these organisms are also common blood culture contaminants, and there is a lack of consensus as to how to interpret CoNS positive results. In the intensive care setting the vast majority of CoNS are resistant to methicillin and thus infants with suspected late onset infection are typically treated with empiric broad spectrum antibiotics that include vancomycin (Rubin 2002). There are concerns regarding unrestricted vancomycin use as a risk factor for the development of resistant organisms, particularly enterococci (HICAC 1995). Its routine use as prophylaxis against nosocomial infection has not been recommended (Craft 1999). Furthermore, the majority of pathogens associated with fulminant late onset sepsis (lethal within 48 hours) have been shown to be gram negative organisms with pseudomonas leading the table (Karlowicz 2000). Therefore, although CoNS are currently the most prevalent pathogens in late onset sepsis in neonatal intensive care units there is a low frequency of associated mortality.

Empiric antibiotic treatment varies between neonatal intensive care units and countries and there are currently no consensus guidelines on the choice of empiric antibiotics. There are also no definitive guidelines on classification of CoNS as true sepsis or contaminant, the removal of indwelling catheters or the duration of antibiotics for late onset sepsis. The definition of late onset sepsis varies between trials and for the purpose of this review we will not include those infants commenced on antibiotics prior to 48 hours of age.