Limited diversity in the gene pool allows prediction of third-generation cephalosporin and aminoglycoside resistance in Escherichia coli and Klebsiella pneumoniae.

Early appropriate antibiotic treatment reduces mortality in severe sepsis, but current methods to identify antibiotic resistance still generally rely on bacterial culture. Modern diagnostics promise rapid gene detection, but the apparent diversity of relevant resistance genes in Enterobacteriaceae is a problem. Local surveys and analysis of publicly available data sets suggested that the resistance gene pool is dominated by a relatively small subset of genes, with a very high positive predictive value for phenotype. In this study, 152 Escherichia coli and 115 Klebsiella pneumoniae consecutive isolates with a cefotaxime, ceftriaxone and/or ceftazidime minimum inhibitory concentration (MIC) of ≥ 2 μg/mL were collected from seven major hospitals in Sydney (Australia) in 2008-2009. Nearly all of those with a MIC in excess of European Committee on Antimicrobial Susceptibility Testing (EUCAST) resistance breakpoints contained one or more representatives of only seven gene types capable of explaining this phenotype, and this included 96% of those with a MIC ≥ 2 μg/mL to any one of these drugs. Similarly, 97% of associated gentamicin-non-susceptibility (MIC ≥ 8 μg/mL) could be explained by three gene types. In a country like Australia, with a background prevalence of resistance to third-generation cephalosporins of 5-10%, this equates to a negative predictive value of >99.5% for non-susceptibility and is therefore suitable for diagnostic application. This is an important proof-of-principle that should be tested in other geographic locations.