Quality of antimicrobial susceptibility testing in the UK: a Pseudomonas aeruginosa survey revisited.

Open Access
Publisher Website
Google Scholar
Abstract
As part of a programme to assess the usefulness of routine antimicrobial susceptibility data as a surveillance tool, we reviewed the results of a national survey of resistance in Pseudomonas aeruginosa, undertaken in 1993. Twenty-four UK laboratories contributed isolates for centralized MIC testing, indicating also their own susceptibility test data. As reported previously (Chen et al. (1995) Journal of Antimicrobial Chemotherapy 35, 521- 34), the rate of false resistance (isolates reported susceptible, but found resistant on MIC testing/all isolates reported susceptible) was 0.6- 8%, according to the antimicrobial and breakpoint. Review showed that this favourable position reflected the fact that >88% of isolates were susceptible to any given antimicrobial and— in most cases— were correctly reported as such. Reporting was more erratic for resistant isolates: for β -lactams and amikacin, isolates resistant at the highest MIC breakpoints were equally likely to be reported as ‘susceptible’ or ‘resistant’; such misreporting was less common with ciprofloxacin and gentamicin but still occurred in 9- 20% of cases. Conversely, up to 73% of the isolates reported as resistant proved to be susceptible at high breakpoints, and up to 44% were susceptible at low breakpoints. Miscategorizations did not reflect failure to detect particular mechanisms but, rather, the fact that MIC and zone breakpoints for P. aeruginosa serve to cut ‘ tails’ of resistant organisms from continuous distributions, not to distinguish discrete populations. In this situation, some disagreement between routine tests and MICs is inevitable, but the frequency at which highly resistant isolates were reported as sensitive is disturbing. For surveillance, we conclude that resistance rates based on routine tests are unreliable for P. aeruginosa. This situation may improve with greater standardization of routine testing, but the continuous susceptibility distributions without discrete resistant and susceptible populations militate against perfect agreement. Despite these deficiencies, routine data should allow trend analysis.