Comparison of Etest with agar dilution for testing the susceptibility of Pseudomonas aeruginosa and other multidrug-resistant bacteria to colistin

Sir, Colistin, introduced for clinical use > 45 years ago, has not been widely prescribed because of an alleged poor activity and high toxicity.1 The increasing prevalence of multidrug-resistant Gram-negative bacteria has stimulated the potential indication of colistin in life-threatening infections. In recent studies, colistin has been considered as an effective and less toxic compound than in older studies, under the conditions of modern patient care and drug monitoring.1,2 Colistin is used parenterally as sodium colistin methanesulfonate, which is less toxic than the locally used colistin sulphate, despite having less favourable pharmacokinetic properties.1 However, colistin-resistant mutants can be easily selected in vitro,3 particularly in Pseudomonas aeruginosa and Acinetobacter baumannii, with a frequency of 10−6 cfu/mL. In the latter species, this could be related to the presence of heteroresistance.1 As already published 38 years ago,4 and occasionally ‘rediscovered’, colistin diffuses poorly in agar, resulting in highly unreliable disc tests; up to 89% of results were categorized as very major discrepancies in a recent study, emphasizing the need for alternative testing methods.5 The Etest is a simple and accurate method for determining antibiotic susceptibility. However, it has not been validated against a large number of strains with acquired colistin resistance: in a recent study with A. baumannii, poor results have been observed for resistant strains.6 The aim of our study was to evaluate colistin Etests versus agar dilution against colistin-susceptible and -resistant clinical isolates. A total of 170 clinical isolates including 31 colistin-resistant strains and 22 in vitro selected mutants were studied. The isolates were from respiratory (41.4%) and urinary tract infections (28.3%), blood cultures (5.2%), skin and soft tissue infections (16.2%) and other sites (8.9%), and from patients treated with colistin or infected with multidrug-resistant bacteria. Bacterial strains included Enterobacteriaceae, P. aeruginosa and other Gram-negative bacilli (Table 1). Agar dilution MICs were determined on Mueller–Hinton agar (Bio-Rad, Marnes La Coquette, France) with adjusted Ca2+ and Mg2+ concentrations according to CLSI recommendations and manufacturer indications, using a Steers replicator device delivering 103–104 cfu/spot. Etest MICs were determined as recommended by AB Biodisk (Solna, Sweden) on the same medium, using an inoculum equivalent to a 0.5 McFarland standard. MICs were read after 18 h of incubation at 37°C in air. Escherichia coli ATCC 25922 and P. aeruginosa ATCC 27853 were tested as controls in five separate determinations. Comparative MIC distributions obtained by agar dilution and Etest Bacterial strains, total (colistin-resistant): A. baumannii, 14 (6); Aeromonas hydrophila, 1; Alcaligenes xyloseoxidans, 4 (2); Burkholderia cepacia, 1 (1); Citrobacter freundii, 11 (4); Enterobacter aerogenes, 6 (5); Enterobacter amnigenus, 1; Enterobacter cloacae, 22 (6); E. coli, 22 (4); Hafnia alvei, 1; Klebsiella pneumoniae, 36 (6); P. aeruginosa, 64 (12); Stenotrophomonas maltophilia, 9 (7). aP. aeruginosa only. Comparative MIC distributions obtained by agar dilution and Etest Bacterial strains, total (colistin-resistant): A. baumannii, 14 (6); Aeromonas hydrophila, 1; Alcaligenes xyloseoxidans, 4 (2); Burkholderia cepacia, 1 (1); Citrobacter freundii, 11 (4); Enterobacter aerogenes, 6 (5); Enterobacter amnigenus, 1; Enterobacter cloacae, 22 (6); E. coli, 22 (4); Hafnia alvei, 1; Klebsiella pneumoniae, 36 (6); P. aeruginosa, 64 (12); Stenotrophomonas maltophilia, 9 (7). aP. aeruginosa only. Because of reproducible results obtained with the reference P. aeruginosa strain ATCC 27853, MICs of ≤4 mg/L were considered to indicate susceptibility. The susceptibilities to colistin of the 170 isolates and 22 mutants are indicated in Table 1. The MICs determined by agar dilution for 108 Enterobacteriaceae were 0.25–1 mg/L (mode 0.5 mg/L) for the susceptible population and 32 to >128 mg/L (mode 32 mg/L) for the resistant population. For the 64 P. aeruginosa isolates, the MICs of colistin were 1–4 mg/L (mode 4 mg/L) for the susceptible population and 8–16 mg/L (mode 8 mg/L) for isolates considered as resistant. When determined by Etest, there was a shift towards 2-fold lower MICs (one dilution) for 37.3% of the isolates and 4- to 8-fold lower MICs (two to three dilutions) for 6.5% of the isolates. Overall, 91.2% of results were in agreement i.e. within a 2-fold dilution. Susceptible and resistant Enterobacteriaceae were clearly delineated with MICs ≤1 mg/L for the susceptible population and ≥32 mg/L for the resistant population, resulting in 100% agreement between the two methods. In contrast, the range of MICs for P. aeruginosa (1–16 mg/L) overlapped the susceptibility breakpoint. The 2-fold shift towards lower MICs and lack of an ‘intermediate’ category resulted in 7 of 12 colistin-resistant P. aeruginosa isolates being misclassified as susceptible by the Etest. This is, however, not significant: a clinically relevant approach has to take into account the site of the infection, the colistin level obtained at this site and the MIC for the strain. Despite this, the 91.2% agreement ( ± a 2-fold dilution) observed in comparison with the reference agar dilution method indicates that the colistin Etest is a useful method for detecting colistin-resistant Gram-negative bacteria. Colistin is more often used in France than in other countries, both locally and parenterally, in part because of the high prevalence of multidrug-resistant bacteria. However, antibiotic susceptibility testing is still performed in many laboratories by the disc diffusion technique, which is highly unreliable for colistin. We strongly recommend that colistin MICs should be determined by Etest for all patients with severe infections caused by multidrug-resistant strains or in cases of clinical failure, and we emphasize the need to transmit these results to the clinician. This study has been supported, in part, by Aventis, France. None to declare.