Background

Prompt treatment of neonatal infection with appropriate antibiotics can be life-saving, however, the availability of effective antibiotics may become limited by the development of resistant organisms.1,–,3 Outbreaks of resistant infections result in closure of neonatal units, with responsible organisms including methicillin-resistant Staphylococcus aureus (MRSA), extended spectrum β-lactamase-producing Gram-negative bacilli (ESBLs), vancomycin-resistant Enterococci (VRE) and Serratia marcescens.4 In a recent UK surveillance study, for example, it was noted that half of all MRSA bacteraemias in children occur in the first month of life.5 The development of antimicrobial resistance is at least partly driven by antibiotic prescribing, particularly of broad spectrum antibiotics. In one US study, intrapartum antibiotic prophylaxis with ampicillin resulted in a significant decline in Group B streptococcus (GBS) invasive disease in very low birthweight babies, but the overall rate of early onset sepsis (EOS) rate was unchanged due to the concomitant increase in Escherichia coli infection.6 This was not observed, however, in a unit where penicillin was the preferred agent for GBS prophylaxis.7 In another study, the routine use of a broad spectrum cephalosporin-based combination was associated with an increase in neonatal colonisation with resistant and unusual organisms, when compared with the use of a narrow spectrum aminoglycoside-based combination.8 Cephalosporins have also been associated with an increased risk of fungal sepsis in very low birthweight babies.9 10 Recent data from Asian neonatal units indicate how difficult the management of neonatal infection can become if multi-resistant pathogens become common.11

Empiric antibiotic use

A recent audit of empiric antibiotic use among UK neonatal units revealed that although nearly 70% chose a narrow spectrum penicillin/gentamicin combination for presumed EOS, 19% used a cephalosporin, either alone or in combination with a penicillin.12 For late onset infections, the choice of empiric antibiotic combinations was much more diverse; 56% used a penicillin/gentamicin or flucloxacillin/gentamicin combination while around 23% used a cephalosporin, either alone or in combination with another antibiotic, most commonly vancomycin. Vancomycin and teicoplanin were also prescribed in 21% and 8% of neonatal units, respectively, in the presence of an indwelling catheter.12

Current epidemiology of neonatal infection

Screens for EOS are undertaken in approximately 10–12% of all babies.13 In one audit of early-onset infection screens, only 13 of 413 (3.2%) babies who were screened for EOS ultimately had any microbiological or clinical and laboratory evidence of bacterial infection.13 They were all treated with antibiotics empirically for at least 2 days. One interpretation of this is that >95% of the babies screened and treated for possible sepsis do not in fact have sepsis and need not have been exposed to antibiotic therapy at all. The most common presenting features of EOS include tachypnoea and associated respiratory signs, temperature instability and feeding difficulties. The difficulty for clinicians is that while EOS is a relatively rare cause of such symptoms, it can be life-threatening if not treated promptly. Being able to accurately differentiate infectious from non-infectious causes, urgently requires improvements in methods of identifying or excluding infection. While few would argue with empiric use of antibiotics for suspected infection, clinicians should also stop antibiotics as soon as possible if subsequent results do not support a diagnosis of infection, and ensure that the antibiotics used are as narrow a spectrum as possible.

Using surveillance systems to identify pathogens and their antimicrobial resistance patterns, as well as engaging with local microbiologists, helps to inform clinicians of what isolates to target and assists clinicians in using antibiotics judiciously. Neonatal infection surveillance systems have not been relatively well developed in the UK until recently. Surveillance of blood stream infections through the Health Protection Agency (HPA), although passive, is able to indicate the range of likely pathogens and their relative frequency as causes of bacteraemia.14 Such surveillance is somewhat limited by lack of clinical details, particularly relevant in interpreting positive cultures with possible commensal organisms. To complement this source, another surveillance network (Neonatal Infection Surveillance Network, NeonIN) was established in 2004 to collect clinical as well as microbiological data on episodes of neonatal infection. Data regarding the incidence, pathogens and antibiotic resistance profiles of infections captured in NeonIN in the 3 years between 1 January 2006 and 31 December 2008 from 12 English neonatal units, have just been published.15

Early onset sepsis

Early onset neonatal sepsis has been variably defined as that occurring within the first 48–72 h, or within the first week of life. The incidence of culture-positive EOS (defined as sepsis in the first 48 h of life) among NeonIN Units is 0.9 per 1000 live births and 9/1000 neonatal admissions. Both the HPA and NeonIN surveillance systems demonstrate that GBS is the predominant bacteria isolated in EOS (50% in neonIN).14 15 This is followed by Gram-negative isolates (25%, predominantly E coli15) and then in similar proportions, other streptococci and S aureus (6% and 5%, respectively15). Listeria monocytogenes is isolated less frequently (0.9–6%14 15 of isolates), but has a significant case death rate, and is an important cause of meningitis. Analysis of antibiotic susceptibility profiles for EO pathogens is possible through both surveillance systems and reveals that the majority (>95%) of organisms causing EOS are susceptible to the two most commonly used antibiotic regimens (benzylpenicillin and gentamicin or amoxicillin and cefotaxime).

Late onset sepsis

Late onset sepsis (LOS) is variably defined as sepsis after the first 48 or 72 h or after the first days or week of life. The incidence of late onset infection (defined as occurring after 48 h of life)15 is approximately 8/1000 live births, affects approximately 7% of neonatal unit admissions and is dominated by coagulase negative staphylococci (CoNS) in 54% of all LOS cases. The majority of CoNS infections are in infants ≤32 weeks gestation (86%) and in extremely low birthweight (ELBW) babies.15 If CoNS are excluded, the majority of LOS isolates (S aureus, E coli, Enterococcus, Enterobacteriacae) are susceptible to the two most commonly used empiric antibiotic combinations (flucloxacillin and gentamicin or amoxicillin and cefotaxime) although susceptibility to flucloxacillin and gentamicin has been found to be higher (84%) than that of amoxicillin and cefotaxime (79%) due to the relatively high resistance rate of Enterobacteriacae to cephalosporins.14 15

Neonatal deaths from infection

An assessment of neonatal deaths attributable to infection provides a different perspective of the burden of infection. Using information from anonymised death registrations in England and Wales over a 3-year period, Depani et al calculated that overall 11.5% of neonatal deaths were attributable to infection.16 Not surprisingly, the most important bacterial pathogen associated with death was GBS followed by E coli, Pseudomonas, CoNS and Klebsiella. When compared to their frequency as causes of bacteraemia this suggests a significant difference in virulence. CoNS is the most frequently isolated pathogen causing LOS, yet appears to be relatively benign when compared to organisms such as Pseudomonas and Klebsiella, which are much less commonly isolated, but disproportionately associated with death. CoNS were only a cause of death in ELBW babies, and even in this group CoNS were an uncommon cause of death.16 The perceived wisdom has been that CoNS infection is also less likely to be associated with poor neurodevelopmental outcome. However, in one study in ELBW babies, for all neurodevelopmental outcomes other than hearing impairment, CoNS infection has been associated with the same risk of neurodevelopmental impairment, as that identified for Gram-negative and other infections,17 suggesting CoNS may be less benign in some babies <1000 g. This information can be useful when considering appropriate empiric therapy for known pathogens in different patient groups.

Recommended antibiotic regimens

Restrictive policies dictate that antibiotic therapy should be with narrow spectrum antibiotics wherever possible and only used when significant infection is likely. Broad spectrum antibiotics should be held in reserve.18 Colonisation of babies, for example, of endotracheal secretions or of skin and mucosal surfaces, without clinical signs of infection, does not warrant antibiotic treatment.18 19

There are no definitive randomised controlled trials of the best antibiotic regimens for the newborn. Each antibiotic has benefits and side effects which should be evaluated every time antibiotics are prescribed. Best practice recommendations based on currently available data are summarised in table 1.

Antibiotic choices for early-onset infection

The empiric combination of benzyl penicillin with an aminoglycoside such as gentamicin provides excellent coverage for UK EOS pathogens while maintaining a relatively narrow spectrum.14 15 A cephalosporin-based combination does not provide significantly better coverage of likely bacteria yet is clearly associated with a broader spectrum of action, and therefore may result in greater potential harm. Furthermore, a combination of cefotaxime/ampicillin as opposed to gentamicin/ampicillin as empirical treatment for EOS has been associated with increased mortality (OR 1.5; 95% CI 1.4 to 1.7).20

Although gentamicin has some activity against S aureus, flucloxacillin remains the best choice antibiotic for methicillin-sensitive S aureus (MSSA). Clinical correlates of S aureus infection should be sought in future studies to enable empiric use of flucloxacillin where this pathogen is more likely. Additionally, when there is a suspicion that clinical progress is suboptimal, consideration should always be given to an empiric change of antibiotic therapy to include a broader spectrum of pathogens such as S aureus. A recent multiple logistic regression analysis revealed that EOS empiric treatment failure could be predicted at 24 h using the following variables: need for vasoactive treatment (OR 2.83 (1.21 to 6.66)); WBC <5000 or >20 000 per mm3 on day 1 (2.51 (1.09 to 5.81)); the ratio of immature to total neutrophils (I/T ratio) >0.2 on day 1 (2.79 (1.10 to 7.11)); the maximum normal I/T ratio value being 0.12–0.16; and if there was persistent thrombocytopenia on day 1.21 Such analyses should be validated in other datasets but have the potential to improve neonatal outcome by ensuring appropriate antibiotics are used as early as possible.

Listeria is rare, and while awaiting cultures is adequately treated with the empirical combination of penicillin and gentamicin.

Disadvantages of aminoglycosides

The disadvantages of using an aminoglycoside include the need to monitor levels and the dosing errors which may occur.22 There is also anxiety regarding gentamicin-induced ototoxicity and sensorineural hearing loss. There are two mechanisms for gentamicin-associated toxicity: the result of sustained high trough gentamicin concentrations and genetically determined ototoxicity. The link with high trough gentamicin levels is based on published case series' from times when extended-interval gentamicin dosing regimens were not in common use. As the bactericidal activity of gentamicin depends on the peak gentamicin concentration, and gentamicin continues to have a bactericidal effect even after this peak has been obtained (the so called ‘post-antibiotic effect’), extended interval dosing regimens have been developed. Such regimens result in high peak levels but there is improved adherence to appropriate trough concentrations (generally accepted as <2 g/l).23 24 Gentamicin continues to be a significant source of medication errors22 and the large number of recommendations on doses and monitoring that exist in the literature and in local guidelines may well explain this. There is an urgent need for a UK-wide, single and standardised approach to neonatal gentamicin dosing and monitoring. Both multiple dose and extended interval dosing regimens are still recommended in the BNF-C and we suggest that this changes to include only an extended interval regimen.

The genetic link between gentamicin and sensorineural hearing loss is also of concern, following the finding that approximately 1:500 of the population carry the mitochondrial DNA mutation m.1555A→G. Carriers of this mutation have permanent and profound hearing loss after receiving aminoglycosides even when drug levels are within the therapeutic range.25 26 For this reason, elective genetic testing is increasingly being considered prior to gentamicin administration, although this has not been evaluated in routine practice especially in the neonatal setting. Further studies are required to evaluate the risk of genetically determined sensorineural hearing loss with neonatal gentamicin use, the practicalities of screening for such mutations and the best alternative antibiotics when gentamicin is contraindicated. There may be a role for ciprofloxacin in this situation. An ongoing European study will provide much needed data in this area.27

Antibiotic choices for late-onset infection

CoNS are the most frequent bacteria isolated from blood cultures in the context of late onset infections. Vancomycin and teicoplanin are the antibiotics of choice for a proven and significant CoNS infection but their excessive use has been associated with the development of VRE infections and of Gram-negative infections.3 28 Targeting empiric use of these agents only to those babies with the highest risk of complicated CoNS infections (eg, resulting in sepsis syndrome, poor neurodevelopmental outcome or death) would greatly minimise their exposure in the neonatal unit. Their use as first-line antibiotics for nosocomial infection should be avoided, as 80% or so, that is, the majority of blood cultures drawn will be negative.29

The majority of the other leading causes of LOS can be appropriately treated by a relatively narrow combination such as flucloxacillin and gentamicin. This includes MSSA, the second most common cause of LO bacteraemia and an organism capable of causing overwhelming infection if not treated early.30 It also includes most of the Gram-negative bacteria, some of which, for example, Pseudomonas, may also have a significant mortality as indicated earlier. In contrast, while the overall coverage is similar to that of flucloxacillin and gentamicin, a cephalosporin alone or in combination with ampicillin may not adequately cover a number of the Enterobacteriaceae.14 15 As with EOS, in situations where clinical improvement is not evident or deterioration is occurring, an empiric change from flucloxacillin to vancomycin or teicoplanin should be considered, together with switching gentamicin for another antibiotic with broader activity against Gram-negative bacteria, for example, piperacillin/tazobactam (tazocin). The use of piperacillin/tazobactam rather than ceftazidime has been associated with a reduction in neonatal Klebsiella infections.31 A combination of vancomycin and gentamicin also provides good Gram-negative and Gram-positive cover but potentially has additive toxicity; this combination should therefore be used with caution.

The bactericidal activity of vancomycin is related to its trough concentration. It is therefore vital that the concentration of vancomycin is maintained well above the minimal inhibitory concentration (MIC) of the organism (eg, at least 3–4 times), at all times during treatment. This is the rationale for developing vancomycin dosing regimens whereby a continuous infusion of vancomycin is given. As many organisms including CoNS have become more resistant over time,32 this level is now set at 10–15 µg/ml rather than 5–10 µg/ml. There is a need for more contemporary data on the vancomycin MICs of relevant pathogens as well as on the best dosing schedules (ie, the use of loading doses and intermittent vs continuous infusion), especially for very premature babies.

Antibiotic therapy should be stopped after 36–48 h if cultures are negative and there are no further signs of infection. It is very unlikely that a culture that becomes positive after more than 48 h in an asymptomatic baby is of clinical significance.29 Furthermore, a 3-day period of incubation is sufficient to detect all clinically relevant infections using the routine BacT/Alert microbial detection system.29 Conversely, once a relevant isolate has been obtained, treatment should be adapted appropriately to ensure that an effective antibiotic with the narrowest spectrum possible is being used.12 18 28 The role of specific neonatal ‘infection rounds’ may be helpful in this context.

Monitoring response to therapy

Antibiotic therapy alone may not clear an infection. If the baby remains unwell, or in the presence of other laboratory indicators including: persisting thrombocytopenia and/or neutropenia, raised C-reactive protein, procalcitonin, or plasma lactate; and always if there is persistence of positive blood cultures, further sets of blood cultures should be taken. As discussed earlier, more work is needed on timely clinical markers that indicate an inadequate response to treatment. Reasons for persistent positive blood cultures include: inadequate antibiotic levels or regimens; resistant organisms, colonisation of indwelling ‘foreign bodies’, for example, long line, umbilical arterial or venous lines; focal infection, for example, necrosis of gut, especially during ‘conservative’ management of necrotising enterocolitis (NEC), abscess formation, osteomyelitis or endocarditis. Evaluation by paediatric surgeons, paediatric orthopaedic surgeons and cardiologists respectively should be sought for such cases.

Apart from repeating blood cultures, further investigations for consideration should include lumbar puncture and suprapubic urine aspirate, checking antibiotic levels, ultrasound of potential foci of infection, skeletal radiographs and echocardiogram. Consideration should also be given to therapeutic options such as optimising antibiotic doses, changing antibiotic regimens or removing indwelling catheters.

The role of adjuvant therapy with immunoglobulins in sepsis remains uncertain but should be revealed in the near future when the INIS trial is completed.33 Haemopoietic growth factors may also be considered under extreme circumstances, but have no place in routine management of sepsis.34

Length of treatment

This will depend on a number of variables such as the type of organism, antibiotic levels that have been achieved, the presence of indwelling catheters and clinical response. There is little published evidence to inform the optimal length of a course of antibiotics for culture-proven infection in neonates, and the following represents the consensus of the authors.

As a general rule, antibiotics should be commenced promptly if there is a possibility of infection, and stopped at 36–48 h in an asymptomatic baby if there is no subsequent clinical evidence of infection, and cultures are all negative.29 This should be the norm, that is, the default position is to stop, not continue antibiotics after a septic screen. Prolonged routine empirical antibiotic therapy (>5 days), among neonates <1000 g birth weight has been associated with an increased risk of death and NEC.35 This association is biologically plausible in view of the evidence linking disordered immune development with antibiotic-induced changes in intestinal microflora36 and pathogenic intestinal bacteria with the development of NEC.

If antibiotics are started on suspicion of infection but cultures are negative, perhaps as a result of prior antibiotics, yet the clinical impression at the start of treatment was that sepsis was very likely, then a longer course may be warranted, for example, 5 days. It is this category of patients where most uncertainty lies, and where non-culture methods of pathogen detection as well as reliable markers of host inflammation offer most potential benefits. Likewise, if there is pneumonia on a chest radiograph, but blood cultures are negative, a 5-day course may also be appropriate. If there are positive blood cultures but negative cerebrospinal fluid (CSF) cultures, treatment should be for a minimum of 10 days. Treatment should be for at least 14 days for S aureus, because of its propensity to seed other tissues, but this decision should be taken in partnership with microbiology or infectious diseases colleagues if possible. For a baby with positive CSF cultures, or a clinical diagnosis of meningitis, then treatment may be required for at least 21 days depending on the organism.37

Osteomyelitis, endocarditis and deep abscesses which are not surgically drained may require several weeks of antibiotic therapy. The length of treatment course may require extension in those with slow clinical and microbiological resolution. The advice of specialists will again be required to manage such situations.

Potential hazards of antibiotics

Antibiotics are increasingly prescribed in the peripartum period, for both maternal and fetal indications. Their effective use undoubtedly reduces the incidence of specific invasive infections in the newborn, such as GBS infection, but a number of potential adverse consequences must also be recognised.

One concern has recently been highlighted by the findings from the 7-year follow-up of the ORACLE trial in which women with spontaneous preterm labour with intact membranes and no overt infection were randomised to receive either erythromycin, co-amoxiclav, both or neither in a factorial design. The study found increased number of children with any functional impairment, OR 1.18 (95% CI 1.02 to 1.37) with the prescription of erythromycin (either with or without co-amoxiclav) and an increased risk of cerebral palsy with the prescription of either or any erythromycin, OR 1.93 (95% CI 1.21 to 3.09), or any co-amoxiclav, OR 1.69 (95% CI 1.07 to 2.67).38 These results demonstrate that the widespread use of antibiotics in late pregnancy may be associated with unexpected long-term consequences in the baby and confirms that antibiotic use is not risk-free.36 39

All antibiotics, particularly broad-spectrum antibiotics, can alter the natural microflora of the patient, particularly in the gastrointestinal tract. This may result in an increase in antibiotic resistance among normal commensal organisms or the emergence of other pathogens. One pathogen may simply be replaced by another pathogen, which itself may be more hazardous, and the total burden of neonatal infection may be unchanged.6 Widespread use of broad spectrum antibiotics within a maternity or neonatal unit will also increase local persistence of resistant organisms and favour opportunistic transmission within the unit.3 26 Such organisms can then persist within hospitals and even in the community. This is demonstrated by a study which aimed to determine the incidence and risk factors for the carriage of multiresistant Enterobacteriaceae strains (MRE; defined as being resistant to three or more classes of antibiotic) and the extent of the persistence of resistant strains following discharge. Sixty-two (50%) of 124 infants had acquired MRE by 2 weeks of postnatal age and 69 (56%) infants had acquired MRE by discharge. At 2 weeks of age, the proportion of babies carrying strains that were resistant to antibiotics were: tetracycline, 79%; amoxicillin, 78%; cephalosporins, 31%; trimethoprim, 20%; piperacillin-tazobactam, 11%; chloramphenicol, 9%; and aminoglycoside, 4%. Babies less than 26 weeks gestation were at higher risk factor of colonisation with MRE at discharge, but not at 2 weeks. Exposure of an infant to a specific antibiotic in the neonatal intensive care unit was not a risk factor for the carriage of a strain resistant to that antibiotic, with the exception of piperacillin-tazobactam.40

A further possible hazard of peripartum antibiotics relates to the potential link with immune dysregulation in later childhood. The developed world has witnessed a substantial increase in the incidence of allergic and autoimmune disease in young children over the past three decades. The suggestion is that immune development becomes abnormal because the naïve immune system is exposed to abnormal bacterial challenge as a result of obstetric practices and inappropriate antibiotic use.36

Longitudinal studies are needed to investigate the potential link between peripartum antibiotic use and immunological health problems in later life. In the meantime, all those involved with caring for the newborn should be aware of these issues; many parents already are, particularly if they already have children with allergies.

The classic microbiological dictum of never using an antibiotic unless one needs to and never using a broad-spectrum antibiotic when a narrow-spectrum one will do, applies more now than ever before.

Suggested action plan

The pipeline for new antibiotics is now worryingly limited. There are very few new antibiotics under development with activity against Gram-negative bacteria and no significantly new Gram-negative antibiotics will be available in the next decade. Evidence from around the world demonstrates that neonatal units are a high risk area for the selection and transmission of multi-resistant organisms. The risk is that recurrent outbreaks of multi-resistant organisms become very difficult to treat. One possible consequence is that the presence of resistant bacteria within neonatal units, which already causes significant disruption to movement of babies between hospitals in neonatal networks could worsen (with serious potential clinical and financial consequences). The balance is switching towards recognition of the need to conserve antibiotics as a finite resource.41 Conservation of antibiotics in hospitals can be performed using some form of Antibiotic Stewardship Programme (ASP).42 This should include some form of regular monitoring of both the total and the spectrum of antibiotics used by conducting a regular point prevalence survey (PPS). In the future, acute hospitals will be encouraged to conduct regular PPS's in all clinical areas. The optimal programme for a neonatal ASP still needs to be developed, but first steps that can be taken now include:

1. Establishing systems for regular infection surveillance of bloodstream infections, either at a local level, for example, by performing an annual audit in collaboration with local microbiology departments and/or through a network such as neonIN (neonin@sgul.ac.uk).

2. Strengthening links with local microbiology and infection control teams and local pharmacists. Having regular (eg, weekly) review of infections and antibiotic use with the focus on stopping antibiotics where possible as discussed above. This could be formalised into a weekly neonatal ‘infection round’, though the evidence base for this has yet to be established.

3. Using a narrow spectrum empiric antibiotic policy (see table 1) and auditing compliance with this.

4. Reinforcing to all staff that the routine policy is for empiric antibiotics to be stopped when blood cultures are negative and auditing compliance with this.

5. Ensuring that continuation of antibiotics after negative blood cultures requires documented justification and a new prescription to be written.

6. Ensuring that the duration of antibiotics to treat an infection episode (whether culture positive or negative) is pre-specified and only prolonged with the writing of a new prescription.

7. Ensuring that the narrowest spectrum antibiotics possible are used for treatment of proven infections and auditing this.

8. Identifying a panel of consultant–decision only antibiotics. We suggest third generation cephalosporins and meropenem, and possibly vancomycin and teicoplanin. Auditing compliance with this and reviewing individual cases where such antibiotics are prescribed.

Putting these measures in place now, while we develop the evidence base for optimal antibiotic prescribing, is the best way of ensuring neonatal infections remain treatable.