Sirs

10.1038/nrmicro1047

A recent opinion1 argues that the rate of isolation of antibiotic-resistant organisms, instead of an estimate of their proportion, provides a better measure of the 'burden of resistance' for public health purposes. The authors also suggest that information derived from proportion-based analyses using existing hospital-based antimicrobial resistance (AMR) surveillance systems might lead to biased estimates of AMR that potentially 'mislead public health professionals' and lack comparability.

We too are aware that existing surveillance systems are imperfect, but argue that not all public health needs will be met by determination of rate-based AMR estimates as proposed by Schwaber et al.1

Between hospitals, and even within the same hospital, differences in patient mix, available services, clinical and sampling practices and laboratory methods for isolation of organisms and susceptibility testing skew AMR estimates. In resource-poor settings, the capacity for AMR surveillance is limited, and data gathering is biased towards resistant isolates. Many AMR surveillance programmes lack valid quality-assurance components. The compounding effects of these deficiencies preclude useful regional, national or international comparisons of AMR data, whether proportion-based or rate-based. The applications of AMR data analyses are therefore in the main limited to a single institution. Pooling large amounts of data from multiple sources does not remedy these defects and at best provides indicative trends that point to a need for more in-depth studies. Wider public health needs require proportion-based or rate-based surveys that are longitudinal instead of sporadic, use constant and continually validated methods, include data on community-acquired pathogens and meet requirements for sample size and representativeness2.

Schwaber et al.1 also allude to a 'clinician's perspective', whereby proportion-based AMR analyses are used to change treatments (for example, for Staphylococcus aureus and penicillins, they suggest a threshold for change of 50%). We contend that properly performed proportion-based AMR estimates expand and help refine this additional public health principle by providing information relevant to both patient care and population disease control. For diseases such as malaria, tuberculosis, gonorrhoea and shigellosis, the determination of the proportion of antibiotic-resistant strains and the probability of success of standard treatments is an integral part of the disease control process and an important public health tool. The highly successful DOTS and DOTS PLUS programmes for tuberculosis control rely on associated and specific AMR surveillance. A recent publication3 linked AMR surveillance to these wider issues of disease control through the use of treatments that interrupt pathogen transmission and eliminate infection reservoirs. Effective treatment of gonorrhoea, determined by AMR surveillance and defined as that single-dose therapy which provides a 95% cure rate, decreases the incidence of gonorrhoea and its complications such as pelvic inflammatory disease, early trimester abortion and ophthalmia neonatorum4. Effective treatment also reduces HIV transmission by eliminating the amplifying effects of gonorrhoea5.

Therefore, we state that the focus of Schwaber et al. could be enlarged to include those specific public health needs that current AMR surveillance systems fail to meet, and that proportion-based AMR surveillance remains one essential component of an integrated programmatic approach to disease control.