INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Cystic fibrosis (CF) is a genetic condition affecting approximately 60,000 individuals in North America and Europe. An important cause of morbidity and mortality in CF is chronic endobronchial infection that is most commonly caused by Pseudomonas aeruginosa (9). Antimicrobial chemotherapy has made an important contribution to increasing the life expectancy of patients with CF (3) and is one of the mainstays of therapy. However, P. aeruginosa isolates from patients with CF frequently have unique phenotypes. They are often multiply resistant and have large amounts of mucoid exopolysaccharide and relatively slow growth rates in the laboratory. These characteristics, thought to be the result of environmental pressure exerted by the milieu in the lungs of patients with CF (4), may adversely affect the performance and interpretation of standardized antimicrobial susceptibility testing.

A consensus conference on CF microbiology was held in May of 1994 (10). This group, composed of microbiologists, infectious disease specialists, and pulmonologists who care for CF patients, evaluated the literature available at the time and recommended the use of disk diffusion susceptibility testing (7) for P. aeruginosa. The use of automated microtiter systems was not recommended because of the perceived inaccuracies of these assays for CF isolates. The current study was undertaken to evaluate those recommendations and to test an additional agar-based stable gradient method for MIC determination (Etest; AB BIODISK, Solna, Sweden). We evaluated the susceptibilities of 597 CF isolates of P. aeruginosa (which included 99 isolates that were tested in two laboratories) to 12 antimicrobial agents, comparing broth microdilution (12) with the disk diffusion method (7) and the Etest.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains. Two hundred fifty isolates were randomly selected from the collection of clinical isolates at the CF Center at the Children's Hospital and Regional Medical Center and the University of Washington in Seattle. An additional 250 isolates were selected from strains sent to the CF Referral Center for Susceptibility and Synergy Testing of Multiply Resistant Organisms at Columbia University in New York, N.Y. Strains were stored frozen at 80°C and checked for purity and mucoid phenotype by plating on MacConkey agar, followed by streaking onto Mueller-Hinton agar to confirm purity and detect pigment production. The following strain definition of P. aeruginosa was used: oxidase positive, catalase positive, growth at 42°C, and pigment production (5). Of the 500 isolates tested, 147 did not meet these phenotypic criteria for P. aeruginosa and thus were further tested by colony hybridization and probing with the exotoxin A gene from P. aeruginosa (13). Seven were found to be exotoxin A negative and were further identified using the API 20 NE strip (bioMerieux Vitek, Inc.) in accordance with the manufacturer's instructions. Two of these seven were not identified as P. aeruginosa and were excluded from further analysis. The 1.4% (7 of 498) proportion of exotoxin A-negative CF isolates of P. aeruginosa is comparable to that reported in the literature (17).

Antimicrobials tested. The following drugs were obtained from Sigma (St. Louis, Mo.) or from their respective U.S. manufacturers: amikacin, gentamicin, tobramycin, ticarcillin, ticarcillin-clavulanic acid (2-µg/ml fixed concentration), piperacillin, piperacillin-tazobactam (4-µg/ml fixed concentration), ceftazidime, aztreonam, ciprofloxacin, imipenem, and meropenem. The ranges of concentrations tested by the reference broth microdilution method were 0.06 to 8 µg/ml for ciprofloxacin; 0.25 to 16 µg/ml for meropenem and imipenem; 0.5 to 64 µg/ml for gentamicin, tobramycin, aztreonam, and ceftazidime; and 2 to 256 µg/ml for all of the other drugs. For disk diffusion, concentrations were in accordance with the National Committee for Clinical Laboratory Standards (NCCLS) recommendations. For Etest, the concentration ranges were 0.002 to 32 µg/ml for ciprofloxacin, meropenem, and imipenem and 0.016 to 256 µg/ml for all of the other agents tested.

Susceptibility test methods. The broth microdilution method (6) was selected as the reference method for this study (12). Drug- and cation-adjusted Mueller-Hinton broth medium-containing trays were custom manufactured by PML Microbiologicals (Wilsonville, Oreg.). Trays were stored at 80°C and allowed to come to room temperature prior to use. Testing was performed in accordance with the recommendations of NCCLS document M7-A4 (6). However, inoculated trays were incubated for 48 h and readings were taken between 18 and 24 h, based on growth in the control well, in accordance with the manufacturer's recommendations, and additionally at 48 h. The MIC endpoint was the lowest concentration of an antimicrobial that completely visibly inhibited the growth in a well.

Retesting of isolates. In the cases in which serious categorical interpretation errors were identified, retesting of the isolates was performed to determine reproducible error. Interpretation errors were classified based on their impact on clinical management. Serious errors included very major errors, defined as susceptible by the test method and resistant by the reference test, and major errors, defined as resistant by the test method and susceptible by the reference method. Minor errors were defined as an intermediate result for either the test or the reference method. For the penicillins, where there is no intermediate category, all errors were either very major or major. All strains with single- or multiple-drug serious errors at the 18- to 21-h incubation were retested in triplicate using the test methodology and the reference broth microdilution test at a third reference laboratory (University of Iowa College of Medicine) by a single experienced technologist. Only retest results are shown in the tables.

Statistical analysis. MICs were converted to a log₂ + 9 values to permit appropriate comparison with each other and with disk diffusion results (millimeters) and entered into the Statistical Analysis System statistical package (SAS Institute, Inc., Cary, N.C.). Pearson correlation coefficients (r values), regression slope equations, and categorical error rates were generated for each antimicrobial agent indexed by susceptibility test method and by mucoid phenotype. A perfect statistical correlation result is 1.00 for the correlation coefficient and/or the slope, and 0.90 (r value) is considered desirable and 0.80 (r value) is considered an acceptable result.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Strain characteristics. The strains tested were highly resistant P. aeruginosa isolates. The interpretive category results of the CF isolates tested by the reference broth microdilution method are listed in Table 1. Overall, only half (47.2%) of the results were in the susceptible range. Tobramycin and meropenem were the most active agents (63.7 and 61.4% susceptible, respectively), while gentamicin (28.6% susceptible) was the least active. MICs clustered near the respective breakpoints accounted for many nonsusceptible results, especially for the penicillins and -lactamase inhibitor combinations, which have no intermediate interpretive category (6, 7, 8).

Interlaboratory variation. The best comparability between the two laboratories performing initial susceptibility testing was for the broth microdilution method, where overall 88.4% of the results for all of the agents were within 1 log₂ dilution and 96.1% were within 2 dilutions (Table 3). The most consistent results were obtained with ciprofloxacin (98.0 and 100% within 1 and 2 log₂ dilutions, respectively); the least consistent were obtained with piperacillin-tazobactam (81.8 and 89.9%, respectively). The disk diffusion test had the least comparable results, with overall 72.2 and 89.3% of results with zones within ±3 and 6 mm, respectively (zone equivalents of ±1 or 2 log₂ dilutions). For this methodology, amikacin results were the least variable (84.4 and 97.9%, respectively) and piperacillin-tazobactam results were the most variable (61.3 and 80.6%). Overall, 83.2 and 91.8% of the Etest results were within 1 and 2 log₂ dilutions, respectively.

Disk diffusion testing. The disk diffusion method results were compared with broth microdilution MICs to examine the frequency of interpretation errors, the correlation coefficient for each tested drug, and the variation of mucoid versus nonmucoid isolates. The analysis of first-run results (data not shown) had an overall 5.5% rate of very major errors, with 6 of 12 agents having unacceptable (>5%) very major errors. When first-run results from incubation for 48 h were compared with those from the standard (18- to 21-h) incubation, there was no clear-cut advantage of one time point over the other.

Etest MIC determination. The Etest method for MIC determination was compared with reference broth microdilution values to examine the frequency of category interpretive errors, the correlation coefficients, and the slope for each drug tested. First-run results (data not shown) for tests incubated for 18 to 21 h had an overall rate of serious errors (very major plus major errors) of 7.1%, with one drug (aztreonam) having a very major error rate of 8.2%. When these results were compared with the results following 48-h incubation, again, no clear-cut advantage was identified for either time point.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This is the most comprehensive study to date to systematically analyze different agar diffusion susceptibility testing methods for P. aeruginosa isolates from patients with CF. Half of the isolates used in this study were selected to be highly resistant, and for a high proportion of them the MICs clustered near the breakpoint concentrations for susceptibility and resistance. In addition, more than one-fourth of the isolates were mucoid. Thus, it was anticipated that such strains would be problematic for the performance of accurate susceptibility testing. In this study, both the disk diffusion method and the Etest performed well for CF isolates, although somewhat better for nonmucoid than for mucoid strains. It had been hoped that more prolonged incubation (up to 48 h) would improve the accuracy of the results for slow-growing strains, but this was not seen.

Inaccurate antimicrobial susceptibility testing for P. aeruginosa isolates from CF patients may have an adverse clinical impact. Many patients with advanced disease are infected with highly drug-resistant strains, and there are few therapeutic options (11). Timely and accurate performance of susceptibility testing may make a significant difference in the therapeutic management of acute pulmonary exacerbation. In addition, laboratories using different techniques may identify different patterns of susceptibility, potentially influencing eligibility for lung transplantation. Finally, reporting of a multidrug-resistant P. aeruginosa strain can affect hospital infection control and isolation policies. A standardized methodology for antimicrobial susceptibility testing is also highly desirable for tracking of resistance patterns in the CF population, especially as new therapeutic agents are developed and traditional antibiotics are delivered to more patients by aerosolization (14).

When the entire population of isolates was examined, both agar disk diffusion and Etest correlation coefficients compared acceptably to the reference broth microdilution methodology. The organisms tested were all from CF patients and were a highly antibiotic-resistant population which mimics those strains that would be problematic. Twenty-seven percent of the isolates were mucoid, which is about half of the percentage (56.7%) reported in a recent study (1). Both tests had poorer correlations for mucoid isolates, although they were still in an acceptable range overall. For disk diffusion, an unacceptable correlation was seen with mucoid isolates for piperacillin, piperacillin-tazobactam, and meropenem. For the Etest, results for mucoid isolates were all within an acceptable range.

Although the correlation with a reference methodology is desirable, the most important information is whether a method of testing can accurately predict susceptibility and resistance. The rate of very major errors (calling an organism susceptible when it is resistant) was, overall, lower with the Etest. The most problematic antimicrobials for the disk diffusion method were ceftazidime, piperacillin, and piperacillin-tazobactam. Overall, major errors (calling an organisms resistant when it is susceptible) for both methods were in an acceptable range (<10%). Because of the lack of an intermediate category for the penicillins, the number of very major and major errors was enhanced for the four penicillin-containing drugs tested (disk diffusion range, 1.9 to 4.9%; Etest range, 4.4 to 6.6%). For both methodologies, the highest rates of serious errors were seen with mucoid isolates. Retesting of strains with minor errors was not performed and might have demonstrated even better performance by each test. However, elevated minor errors were expected due to the clustering of MICs for the CF isolates near the respective breakpoints of the drugs tested.

An important finding in this study was that, overall and for each antimicrobial tested, mucoid isolates in this collection were more susceptible than nonmucoid isolates. There is some controversy in the literature about whether the mucoid phenotype results in increased resistance. A recent study of inhaled-tobramycin therapy examined the susceptibility of 1,240 CF isolates (710 mucoid, 530 nonmucoid) and found that for all seven of the drugs tested, mucoid isolates were more susceptible (14). Several other studies also suggest that mucoid isolates are more susceptible to ciprofloxacin (2) and to the aminoglycosides (16) than are nonmucoid isolates. Another study found that the MIC for 90% of the isolates tested (MIC₉₀) was higher for mucoid versus nonmucoid isolates for ceftazidime, piperacillin, and amikacin but not different for gentamicin, tobramycin, and imipenem (15). However, the concept of increased resistance in mucoid isolates is prevalent in the literature (4) and in clinical microbiology lore.

The results of this study lead to several recommendations regarding susceptibility testing of CF isolates of P. aeruginosa. Disk diffusion and Etest performed acceptably compared to the reference broth microdilution methodology, although both methods performed better for nonmucoid than for mucoid isolates. While we tested only CF isolates, these results might also apply to multiply resistant strains of P. aeruginosa from non-CF patients. Questions that remain with regard to the performance of susceptibility testing of CF isolates include the applicability of the results for P. aeruginosa to other nonfermenting gram-negative bacilli (i.e., Burkholderia cepacia, Stenotrophomonas maltophilia, and Achromobacter xylosoxidans) and what other modalities might improve our ability to test mucoid isolates. Regardless, these two assessed methods were observed to be acceptable for use as routine, practical methods for testing of CF isolates of P. aeruginosa.