Arch Dis Child 98:90-91 doi:10.1136/archdischild-2012-303111
  • Editorials

Combating antibiotic resistance: the war on error

  1. David Andresen1,2
  1. 1Department of Infectious Disease & Microbiology, Children's Hospital at Westmead, Westmead, New South Wales, Australia
  2. 2Discipline of Paediatrics & Child Health, University of Sydney, Sydney, New South Wales, Australia
  1. Correspondence to Professor David Isaacs, Department of Infectious Disease & Microbiology, Children's Hospital at Westmead, Westmead, NSW 2145, Australia; david.isaacs{at}
  • Received 26 October 2012
  • Revised 10 November 2012
  • Accepted 12 November 2012
  • Published Online First 12 December 2012

The treatment of an infection is often described in military terms: the war on microbes, the war on infections, or the battle against antimicrobial resistance. We believe the chance of winning the ‘war on antimicrobial resistance’ by escalation is about as likely as the chance of winning the ‘war on terror’. Our war should be against human error, particularly the erroneous belief that new drug discoveries will be the solution to resistance development. We should de-escalate rather than escalate antibiotic use, invest in diagnostics, ensure our hospitals do not act as ‘resistance amplifiers’, and try to prevent infections.

Downie et al 1 describe Staphylococcus aureus, Escherichia coli and Klebsiella species as the most common causes of community-acquired neonatal and infant bacteraemic infection in developing countries. The importance of Klebsiellae is surprising. This genus is classically associated with nosocomial infections. Some of the primary studies in this review may not have separated nosocomial from true community-acquired sepsis adequately, particularly the largest study by Sharma et al, from India. Nonetheless this meta-analysis of in vitro evidence1 shows that the WHO recommended antibiotic regimen for neonatal sepsis, penicillin and gentamicin, covers only just over half the isolates. The alternative of a third generation cephalosporin, often used because it is cheap and ‘broad spectrum’, does not improve coverage.

Most antibiotics are derived from naturally occurring microbial products. Environmental bacteria found in caves which had been isolated from the outside world for 4 million years have been shown to possess multiple resistance genes to modern antibiotics.2 These genes have evolved to protect them against other organisms and their products during millennia of co-existence and competition for ecological niches. This resistance meta-genome (‘meta-resistome’) is a pool of genes in commensal and free-living bacteria. It represents a rich resource on which bacterial pathogens can draw when challenged by novel antimicrobial agents.3

Antibiotic resistance is an example of Darwinian selection. Following random gene acquisition events, the expression and transmission of resistance genes in pathogenic bacteria is selected by human use and misuse of antibiotics, a process we have called ‘unnatural selection’. Before penicillin was widely used, all S aureus were sensitive to penicillin, but penicillin use rapidly selected for β-lactamase- (penicillinase-) producing strains. Antibiotic resistance has been a major clinical problem ever since.

Gram negative bacilli have a great diversity of resistance mechanisms, and strains that produce powerful extended spectrum beta-lactamases and carbapenemases have emerged worldwide. The genes that code these bacterial self-defence enzymes are often co-located on mobile genetic elements with other resistance determinants. This co-location of diverse resistance genes provides a mechanism by which use of an antimicrobial of one class may select for bacteria resistant to unrelated agents. Worryingly, these highly resistant Gram negative organisms are a major cause of community-acquired early-onset neonatal sepsis in parts of India4 and probably elsewhere in Asia. While resistant bacteria are rarely more virulent, when widespread (or perceived to be so) they often necessitate the use of new, expensive antibiotics, which in turn drives further resistance.5

The reasons for the selection of highly resistant community organisms are complex. Community bacterial resistance rates are not uniform, and probably reflect a myriad of determinants such as sanitation, climate, over-the-counter drug availability, veterinary antimicrobial usage, migration patterns and chance events such as the introduction of successful pathogenic bacterial strains. Once bacterial resistance has emerged, either by gene acquisition from the meta-resistome or by de novo mutations, it is frequently amplified by modern healthcare practices. The healthcare milieu provides a highly efficient transmission setting for both resistant bacteria and resistance genes. Vulnerable patients are housed in close proximity, under dense antimicrobial selection pressure, cared for by well-meaning professionals who may adhere poorly to basic infection control measures.

From a Darwinian perspective, bacteria with resistance genes only have a strong selective advantage in an environment where antimicrobial use is common. The prolonged use of broad-spectrum antibiotics in hospital, notably third generation cephalosporins, extended-spectrum penicillins (piperacillin-tazobactam and ticarcillin-clavulanate) and carbapenems, provides an environment for ‘bacterial evolution in fast-forward’. This may be compounded by sub-optimal diagnostic laboratory support. A clinician who cannot or does not rely on laboratory testing, particularly susceptibility results, lacks the confidence to de-escalate their patients’ antibiotics appropriately, even when a susceptible pathogen is identified. Clinicians may also not trust negative culture results. When babies remain sick, there is a tendency to add ever more broad-spectrum antibiotics, even in the absence of proven sepsis, a practice dubbed ‘spiralling empiricism’.

Another consequence of poor diagnostic support is that units are unable to create summary antimicrobial susceptibility data for local community- and hospital-acquired bacterial infections, invaluable data which would describe the likely success of a given antimicrobial regimen in that geographical location, at that time.

The emergence of multi-resistant organisms in the community casts doubt on the value of the WHO continuing its valiant attempts to develop one-size-fits-all recommendations for suspected neonatal sepsis in developing countries. It is becoming increasingly apparent that empiric antibiotic regimens should be based on the resistance patterns of the local organisms causing neonatal sepsis in a region or institution. Based on the accompanying data,1 the WHO might recommend the combination of an agent with activity against S aureus such as flu/di/cloxacillin or cephalothin (or vancomycin in settings where methicillin-resistant S aureus is very common), with an aminoglycoside with activity against local Gram negatives (eg, gentamicin or amikacin). This would be preferable to third generation cephalosporins, extended-spectrum penicillins or carbapenems on the grounds of spectrum, cost and resistance potential. The argument that cephalosporins are superior to aminoglycosides for Gram negative meningitis1 is not supported by existing clinical data.

Looking at the bigger picture, what is to be done about antibiotic resistance? There is evidence that antibiotic resistance patterns can be reversed by using narrow-spectrum antibiotics instead of broad-spectrum antibiotics in the neonatal unit6 and in the community.5 When local resistance patterns necessitate the empiric use of broad-spectrum antibiotics for neonates with suspected sepsis, these should be stopped after 2–3 days if there is no confirmed sepsis. Prolonged antibiotic use selects for resistant organisms, and is associated with increased mortality and increased risk of necrotising enterocolitis in extremely preterm neonates,7 presumably by altering essential intestinal flora. Although an educational approach would be preferable, reducing the inappropriate use of antibiotics in the community may require national legislation.5 Local, regional, national and international resistance surveillance programmes using robust laboratory and epidemiological methods are critical to monitor the success of resistance prevention measures, and to inform empiric therapy recommendations. Finally, many neonatal infections in developing countries can be prevented through simple, cheap measures such as breast-feeding, umbilical cord care, kangaroo care and hand hygiene.8 We should also investigate potential preventative strategies related to feeding, the physical environment, infection control, immunomodulatory agents, probiotics and vaccines.

Bacteria already possess resistance genes. Escalation of antibiotic use by increasing the use of current broad-spectrum antibiotics and by developing newer, more powerful agents is doomed to failure because it will inevitably select for strains carrying resistance genes to these drugs. We cannot win the war against antimicrobial resistance. Instead we need to win the peace, by using existing antibiotics responsibly and by preventing infections through improved implementation of known preventative strategies and by research into novel preventative approaches.8


  • Funding None.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.


Relevant Articles