INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The clinical significance of anaerobic bacteria and their increasing resistance to antimicrobials have increased the importance of susceptibility testing (3-7, 14). Although there is some debate over the need for susceptibility testing of current clinical isolates, there is little doubt that at least on some occasions, susceptibility results will be of clinical value for practitioners dealing with anaerobic infections. The National Committee for Clinical Laboratory Standards (NCCLS) recommends susceptibility testing in particular situations (e.g., of isolates from brain abscess, endocarditis, joint infections, prosthetic devices, or vascular grafts; from persisting or recurring bacteremia; or from patients who do not respond to empirically chosen chemotherapy) (13). The NCCLS reference agar dilution method is not convenient for testing individual isolates against a large battery of antimicrobial agents. As disk diffusion test results for anaerobes do not correlate with those of the reference method, alternative methods are required by laboratories involved routinely in antibiotic susceptibility testing. The ATB ANA and E test are preferred by most microbiologists. The first ATB ANA strip was marketed with a system containing two fixed antibiotic concentrations and 14 pairs of wells (non-MIC format). In 1993, the ATB ANA became a system using 25 single concentrations. The choice of antimicrobial agents and the single breakpoint concentration relied to a large extent on the NCCLS information supplement (1991) for the susceptibility testing of anaerobes (12). However, the NCCLS has recently introduced an intermediate category; thus, clinical categorization of anaerobes is determined by ATB ANA with two fixed concentrations for most antibiotics. The remaining problem comes from the lack of worldwide agreement on antibiotic breakpoints. French-speaking countries follow the recommendation of the Committee of the Antibiogram of the French Society of Microbiology (CA-SFM) (1), even if there are no specific breakpoints for anaerobes, while other countries either have their own values or apply the NCCLS ones. Thus, for some antibiotics more than two concentrations are proposed, with clinical categorization of strains according to both French and U.S. criteria. The purpose of this study was first to evaluate the performance of the ATB ANA at the concentrations contained in the system. In addition, the interpretation of results according to both CA-SFM and NCCLS breakpoints offers new information that could in the future contribute to the setting up of worldwide antibiotic breakpoints for anaerobes.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. Tests were performed on 200 anaerobes isolated from clinical specimens (Table 1) and identified by classical procedures (17). Of these, 191 were taken arbitrarily. Five strains of Bacteroides fragilis from our collection were added for their resistance mechanisms: three had a carbapenemase, one of these also had low-level metronidazole resistance (MIC, >4 mg/liter), and two other strains were resistant to co-amoxiclav but not to imipenem. Finally, four quality control strains were added: B. fragilis ATCC 25285, Bacteroides thetaiotaomicron ATCC 29741, Clostridium perfringens ATCC 13124, and Eubacterium lentum ATCC 43055. 

Determination of antibiotic MICs. Version 96 of the ATB ANA device was manufactured to correspond with the breakpoints values of NCCLS approved standard M11-A3 (13) with its M100-S6 supplement. At the time we started this evaluation, Wilkins-Chalgren medium was the recommended medium. MICs of amoxicillin, co-amoxiclav, piperacillin, ticarcillin-clavulanic acid, cefoxitin, cefotetan, imipenem, clindamycin, and metronidazole were determined by a reference agar dilution method (NCCLS M11-A3) on Wilkins-Chalgren agar medium (Unipath, Dardilly, France) (21). The final inoculum was approximately 10⁵ CFU per spot of inoculation. Thus, our method was not exactly that currently recommended by the NCCLS, as we used Wilkins-Chalgren medium instead of brucella blood agar, although the NCCLS document indicates that the two media have equivalent performances for all of the antibiotics used here except ticarcillin.

API ATB ANA. The ATB ANA system is a freeze-dried panel with large wells. The ATB ANA strip consists of 16 pairs of cupules. The first pair does not contain any antibiotic and serves as a positive growth control. The next 12 pairs contain antibiotics at two concentrations (corresponding to NCCLS M11-A3 M100-S6 breakpoints): benzylpenicillin, 0.5 and 2 mg/liter; amoxicillin, 2 and 4 mg/liter; co-amoxiclav, 4/2 and 8/4 mg/liter (throughout this paper, for combination drugs the pair x/y refers to the concentrations of the two drugs in the combination); piperacillin, 32 and 64 mg/liter; ticarcillin-clavulanic acid, 32/2 and 64/2 mg/liter; piperacillin-tazobactam, 32/4 and 64/4 mg/liter; cefoxitin, 16 and 32 mg/liter; cefotetan, 16 and 32 mg/liter; imipenem 4 and 8 mg/liter; clindamycin, 2 and 4 mg/liter; chloramphenicol, 8 and 16 mg/liter; and metronidazole, 8 and 16 mg/liter. Three wells, containing amoxicillin at 16 mg/liter, co-amoxiclav at 16/2 mg/liter (a fixed concentration of 2 mg of clavulanic acid per liter in France), and metronidazole at 4 mg/liter, were added when CA-SFM and NCCLS breakpoints for resistance were not identical. We conformed strictly to the recommendations of the manufacturer, as follows. Colonies from Columbia blood agar (after 24 to 48 h of growth) were picked up with a swab and introduced into the suspension medium to produce a turbidity to match the McFarland no. 3 standard (9 × 10⁸ CFU/ml). Two hundred microliters of this suspension was introduced into 7 ml of antibiotic S medium, and 135 µl was further delivered with an automatic pipette (bioMérieux) into each well of the ATB ANA device. Incubation was carried out in an anaerobic chamber with an atmosphere of 85% N₂, 10% H₂, and 5% CO₂. Unless adequate growth is achieved, susceptibility testing cannot be done. The device was read visually by two well-trained technicians as follows: susceptible, no growth; intermediate, growth only at a low concentration; and resistant, growth in both wells of the pair.

Comparison between methods. For all of the strains and from the individual MICs, a clinical categorization (susceptible, intermediate, or resistant) was made according to the different CA-SFM or NCCLS breakpoints (Table 2). These results were compared with those obtained with the ATB ANA device, and errors are defined as follows: minor errors, strains were intermediate by one method and susceptible or resistant by the other; major errors, strains were resistant by ATB ANA, and susceptible by the reference method; very major errors, strains were susceptible by ATB ANA and resistant by the reference method. Category agreement means that strains are classified in the same category by both methods; essential agreement means that there were minor errors only.

Medium substitution. For 40 strains that failed to grow in the ATB S medium, we decided to modify the protocol by using 7 ml of a homemade West-Wilkins broth (20).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Detection of resistant strains and growth by ATB ANA. With the use of NCCLS or CA-SFM breakpoints, the ATB ANA device detected all strains that were resistant to imipenem (3 strains) and piperacillin (27 strains), 32 of 33 clindamycin-resistant strains, and 2 of 2 B. fragilis strains that were resistant to metronidazole at a low level. By contrast, the ATB ANA device failed to detect resistance to co-amoxiclav and ticarcillin-clavulanic acid for four and two strains of the B. fragilis group, respectively. These four strains had a rare resistance mechanism (lack of porin and/or overproduction of the chromosomal -lactamase); they were susceptible to imipenem and not to the combinations of penicillin and -lactamase inhibitor (7).

Results according to NCCLS breakpoints. Category and essential agreements between the two methods were observed for 94.9 and 98.7% of the results, respectively (Table 3). Among the B. fragilis group, errors were more frequent, especially for non-B. fragilis species; 42 of 50 minor errors were due to the clustering of MICs around the breakpoints. For all anaerobes (Table 3), most minor errors (40 of 68) were related to cefoxitin and cefotetan, and category agreement for imipenem was 100%.

Results according to CA-SFM breakpoints. Category and essential agreements between the two methods were observed for 94.3 and 99.1% of the results, respectively (Table 3). For gram-negative bacilli, the more frequent minor errors were related to the clustering around the breakpoint with either amoxicillin (MICs of 16 or 32 mg/liter and breakpoint of 16 mg/liter) or cefotetan (MICs of 32 or 64 mg/liter and breakpoint of 32 mg/liter). Sixty-nine of 87 minor errors were observed within the B. fragilis group. For the nine antibiotics tested, we did not detect any very major discrepancy among gram-positive anaerobes.

Differences according to type of antibiotic breakpoints. The numbers of very major errors were similar (4.2 versus 4.6%), but there were more frequent minor errors (4.8 versus 3.8%) and fewer major errors (0.4 versus 0.8%) when CA-SFM breakpoints were used.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Evaluation of the ATB ANA method. The overall percentages of errors (5.7 and 5.1%) nearly met the 5% limit set by Sherris and Ryan (16) for aerobes. A higher tolerance level would be expected for anaerobes, given the problems inherent in preparing correct inocula and limited growth. Their suggested acceptable level of very major errors (1.5% for all individual species tested) was not exceeded. The other criteria proposed by Metzler and Dehaan (11) (false susceptibility of <1% and false resistance of <5% for all tests), by Jorgensen (10) (false susceptibility plus false resistance of <7%), and by Thornsberry and Gavan (18) (complete category agreement of >90% and false susceptibility plus false resistance of <5%) are met by using the ATB ANA device. The Food and Drug Administration (FDA) has established minimal performance characteristics to assess antimicrobial susceptibility devices (19). These guidelines indicate that category agreement (applied to devices using category result formats) should be >90%, major errors should be <3% for susceptible strains, and very major errors should be 1.5% for resistant strains. Only the last criterion was not satisfied in this study. Of 10 very major errors observed among the 218 (CA-SFM) or 238 (NCCLS) resistant strains (Table 4), half came from the two strains of our collection that were resistant to both co-amoxiclav and ticarcillin-clavulanate but susceptible to imipenem. It is doubtful that any method could satisfy this last criterion when anaerobes are involved. Our results are somewhat better than those of studies performed with an older version of the ATB ANA device (8, 9, 22). The prevalence of both resistant and intermediate strains may greatly influence the evaluation of a new device. In our present study, strains were classified 372 (CA-SFM) or 320 (NCCLS) of 1,800 times as either intermediate or resistant. For some antibiotics (imipenem and metronidazole) to which resistance was rare, it was necessary to add some resistant strains from our collection. For Prevotella and Fusobacterium we were unable to find metronidazole-, imipenem-, or co-amoxiclav-resistant strains, while Veillonella and C. perfringens strains were susceptible to all of the investigated antibiotics. The performance of the ATB ANA device with the nonmanufactured West-Wilkins broth (data not shown) was identical to that with the ATB S medium.

Differences according to type of antibiotic breakpoints. There were 87 (CA-SFM) versus 68 (NCCLS) minor errors. The striking differences came with amoxicillin (27 versus 6 strains), mostly within the B. fragilis group (22 versus 4 strains), and were associated with clustering around the French breakpoints, since for 20 strains amoxicillin MICs were equal to 16 or 32 mg/liter. There were also 6 (CA-SFM) versus 12 (NCCLS) major errors observed within the B. fragilis group and related strains designated resistant to cefotetan and piperacillin by ATB ANA; the differences are attributed to the fact that in France the intermediate zone is wider.

Need and suggestion for new antibiotic breakpoints. Comparison of the MIC distribution and breakpoints suggests that both the French and U.S. versions have useful features (Table 5). For gram-negative anaerobes the NCCLS breakpoints for amoxicillin have many advantages: (i) the amoxicillin breakpoint of 2 mg/liter may successfully separate Prevotella and Fusobacterium strains into two groups in relation to -lactamase production (2) (MIC of 1 mg/liter if negative; MIC of 4 mg/liter if positive), and (ii) at the concentration of 8 mg/liter, nearly all strains of the B. fragilis group are reported as being resistant to aminopenicillins. However, as the NCCLS does not propose breakpoints for gram-positive anaerobes, we suggest using the French ones. The 2-mg/liter concentration of amoxicillin is inadequate for gram-positive anaerobes and will classify strains of Peptostreptococcus, Clostridium difficile, and Eubacterium as intermediate, while the corresponding infections will respond to treatment at normal dosages, even though the amoxicillin MIC is equal to 4 mg/liter. Some strains of Peptostreptococcus that are reported as being resistant to amoxicillin (MIC, 8 mg/liter), will respond to treatment at a higher dosage even though the amoxicillin MIC is equal to 8 or 16 mg/liter.