INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Automatic or semiautomatic commercial systems for bacterial identification and susceptibility testing were introduced in clinical microbiology laboratories more than 20 years ago (1, 11). These systems are specifically designed to allow reliable bacterial identification by using a number of biochemical tests and MIC determinations that are interpreted according to the susceptibility and resistance criteria established by different committees (7). Most of these systems are highly automated, particularly for MIC determinations and interpretations. The final report should offer an acceptable accuracy and should reproduce the values obtained by reference methods (4, 7, 9).

Bacterial identification and susceptibility testing systems vary in the methods that they use to detect bacterial growth and/or to determine endpoints. Either turbidimetric monitoring of bacterial growth and fluorometric detection of the fluorescent indicator or the hydrolysis of fluorogenic substrates is extensively used (7). In contrast, image analysis technology has rarely been applied to these systems and is limited to use with certain devices that analyze inhibition zones obtained by disk susceptibility test methods (3, 10).

The Wider system (Francisco Soria Melguizo, S.A., Madrid, Spain) is a newly developed computer-assisted image-processing device. With the assistance of a video camera it recovers a complete image of commercial microdilution panels used for bacterial identification and susceptibility testing. After image digitization, the Wider system automatically generates the bacterial name and the susceptibility profile, which depend on the analysis of growth and color changes in the identification wells and the interpretation of growth parameters in the susceptibility testing wells, respectively. We report here the results of an evaluation of the newly developed Wider system, which has been adapted to read MicroScan panels (Dade-MicroScan, West Sacramento, Calif.). This work was specifically designed to study the accuracy of this system as a routine tool in clinical microbiology laboratories. Moreover, the susceptibility testing performance of this instrument was also determined with isolates of the family Enterobacteriaceae with known resistance mechanisms.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial isolates. A total of 244 fresh bacterial clinical isolates, prospectively and consecutively collected in our hospital during January and February 1999, were studied. They included 138 isolates of the family Enterobacteriaceae, 25 nonfermentative gram-negative rods (NFGNRs), 51 Staphylococcus spp., 2 Micrococcus spp., 22 Enterococcus spp., and 6 -hemolytic streptococcal isolates. Moreover, 100 clinical Enterobacteriaceae isolates with known resistance mechanisms were also included (Table 1). Prior to identification and susceptibility testing, the organisms were subcultured twice onto 5% sheep blood agar plates.

View this table: [in this window] [in a new window]   TABLE 1.   Enterobacteriaceae with known -lactam-resistance mechanisms

Wider system. The Wider system is basically composed of a reader module assisted by a data analysis module. The reader module is an illuminated chamber with a digitizing video camera that completely reflects the image of a commercial tray used for bacterial identification and susceptibility testing. The only action required by the operator is manual insertion of the trays into the reader module with the assistance of a special support. The rest of the process is computer controlled. The digitized image is analyzed by the Wider system's software in the data analysis module. A clear image appears on the computer screen within 5 s. The software automatically detects the type of panel, assigns identification probability scores as a result of the analysis of growth and color changes in the identification wells, and identifies each isolate by comparing the biochemical profile with the profiles in the software database. Moreover, growth parameters in susceptibility testing wells are analyzed in comparison with those in positive and negative control wells. The MIC of each antibiotic is defined as the lowest concentration with the absence of bacterial growth. For categorization purposes, MICs are interpreted by using either the guidelines of the National Committee for Clinical Laboratory Standards (NCCLS) (20) or those from the Spanish Antibiogram Committee (Mesa Española para la Normalización de la Susceptibilidad y Resistencia a los Antimicrobianos [MENSURA]) guidelines from the Spanish Antibiogram Committee (2).

Bacterial identification and susceptibility testing with the Wider system. Wider system 6W and 3W panels containing lyophilized antibiotics and substrates were used for bacterial identification and susceptibility testing for gram-negative and gram-positive bacteria, respectively. Dade-MicroScan manufactured these panels, which are similar to those used in the overnight WalkAway system (1). Biochemical tests used for both gram-negative and gram-positive organism identification are identical to those included in the WalkAway MicroScan panels. Moreover, the biochemical identification database is the same as that used in the WalkAway system. In addition, Wider 5W panels, which only have antimicrobial agents for susceptibility testing, were also used to assay all isolates. The antibiotics used in the panels (concentration ranges) are as follows: for the Wider panel for gram-negative organisms (reference 6W), amikacin (4 to 16 µg/ml), amoxicillin (4 to 16 µg/ml), amoxicillin-clavulanate (4/2 to 32/16 µg/ml), cefazolin (2 to 16 µg/ml), cefotaxime (0.12 to 8 µg/ml), cefoxitin (4 to 16 µg/ml), ceftazidime (0.5 to 16 µg/ml), ceftazidime-clavulanate (1/4 and 8/4 µg/ml), cefuroxime (1 to 16 µg/ml), ciprofloxacin (0.12 and 1 to 4 µg/ml), fosfomycin (8 to 32 µg/ml), gentamicin (2 to 8 µg/ml), nalidixic acid (4 and 16 µg/ml), nitrofurantoin (64 µg/ml), norfloxacin (1 and 4 µg/ml), ticarcillin (16 to 64 µg/ml), trimethoprim-sulfamethoxazole (2/38 to 4/76 µg/ml), and tobramycin (2 to 8 µg/ml); for the Wider panel for gram-positive organisms (reference 3W), amikacin (4 to 16 µg/ml), amoxicillin-clavulanate (4/2 to 32/16 µg/ml), ampicillin (0.5 to 16 µg/ml), cefazolin (2 to 4 µg/ml), cefotaxime (0.06 to 4 µg/ml), cefuroxime (0.5 to 2 µg/ml), chloramphenicol (8 µg/ml), ciprofloxacin (0.5 to 4 µg/ml), clindamycin (0.5 and 2 µg/ml), erythromycin (0.12 and 0.5 to 2 µg/ml), fosfomycin (32 to 64 µg/ml), gentamicin (2 to 8 and 500 µg/ml), oxacillin (1 to 4 µg/ml), penicillin (0.06 to 8 µg/ml), rifampin (0.5 and 2 µg/ml), streptomycin (1,000 µg/ml), teicoplanin (1 to 16 µg/ml), trimethoprim-sulfamethoxazole (1/19 to 2/38 µg/ml), and vancomycin (1 to 16 µg/ml); for the supplementary panel (reference 5W), aztreonam (0.12 to 16 µg/ml), cefepime (0.12 to 16 µg/ml), ceftriaxone (0.12 to 16 µg/ml), chloramphenicol (1 to 8 µg/ml), colistin (2 to 4 µg/ml), imipenem (0.125 to 16 µg/ml), meropenem (0.12 to 16 µg/ml), ofloxacin (0.06 to 8 µg/ml), piperacillin (8 to 64 µg/ml), piperacillin-tazobactam (8/4 to 64/4 µg/ml), sulbactam (1 to 8 µg/ml), tetracycline (1 to 16 µg/ml), and trovafloxacin (0.06 to 4 µg/ml).

Reference methods. API galleries (API 20E, API 20NE, API Staph, and API Strep; BioMerieux, SA, Marcy-l'Étoile, France) were used as the reference tests for bacterial identification. Conventional biochemical tests (17) were also performed when discrepancies were observed. The MIC results obtained with the Wider system were compared with those obtained by the reference broth microdilution method described by NCCLS (19). The final inoculum concentration was 5 × 10⁵ CFU/ml. Antimicrobial powders were obtained from their respective manufacturers.

Quality control. Quality control was assured by running every day the organisms recommended for this purpose by NCCLS (20): Escherichia coli ATCC 25922 and ATCC 35218, Staphylococcus aureus ATCC 29213, Enterococcus faecalis ATCC 29212, and Pseudomonas aeruginosa ATCC 27853. The reproducibility of the MIC readings was also analyzed with the American Type Culture Collection (ATCC) strains. These strains and Proteus vulgaris ATCC 13315, Klebsiella pneumoniae ATCC 29665, Klebsiella oxytoca ATCC 49131, Salmonella enterica subsp. arizonae ATCC 12323, Enterobacter cloacae ATCC 13047, Stenotrophomonas maltophilia ATCC 13637, Staphylococcus epidermidis ATCC 12228, and Listeria monocytogenes ATCC 19112 were run during the evaluation as quality controls for bacterial identification.

Analysis and accuracy of results. The identification scores provided by the Wider system software were not considered in the analysis; instead, the final identification results were used. A "correctly identified" category was established when identical identifications at the species level were provided by the Wider system and by the reference method. The "partially identified" category meant that the strain identifications were coincident only at the genus level, and the "incorrectly identified" category meant that the genus assigned by the Wider system was different from that obtained assigned by the reference method.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Organism identification. Table 2 shows the performance of the Wider system for bacterial identification. A total of 244 isolates routinely obtained from different clinical sources were tested. The Wider system correctly identified 97.5% of these bacterial clinical isolates: 99.3% of the isolates of the family Enterobacteriaceae, 92.0% of the NFGNRs, and 96.3% of the gram-positive organisms. Misidentifications at the genus level (incorrect identification) were limited to only two NFGNRs. In addition, one enterobacterial isolate and three staphylococcal isolates were partially identified (misidentification to the species level). It must be stressed that a Salmonella sp. isolate and an S. aureus isolate were incorrectly identified as S. enterica subsp. arizonae and a coagulase-negative staphylococcus, respectively. Bacterial strains for identification quality control were always correctly identified, including S. enterica subsp. arizonae ATCC 12323 and S. aureus ATCC 29213. 

Susceptibility testing of routine isolates. A total of 3,719 organism-antimicrobial agent combinations were analyzed: 2,484 for isolates of the family Enterobacteriaceae, 300 for NFGNRs, and 893 for gram-positive isolates. The overall essential agreement in MICs (±1 log₂ dilution) for all these organism-antimicrobial agent combinations was 95.6%; 1.6% of the results with the Wider system were 2 or more dilutions lower than the reference MICs, and 2.8% of the results were 2 or more dilutions higher than the reference MICs.

Susceptibility testing of Enterobacteriaceae with well-characterized -lactam resistance mechanisms. A total of 800 organism-antimicrobial agent combinations were analyzed. The essential agreement of susceptibility testing of isolates of the family Enterobacteriaceae with known -lactam resistance mechanisms with the Wider system and by the standard microdilution method was 94.8% (Table 6), which is slightly lower than that obtained with routine isolates of the family Enterobacteriaceae (97.4%). The highest essential agreement was observed with chromosomal AmpC -lactamase-producing isolates (97.7%) and permeability mutants (98.4%). On the contrary, the lowest essential agreement was obtained with chromosomal class-A -lactamase-producing isolates (89.1%). The analysis of the different antimicrobial agents tested revealed that more than 97.0% essential agreement was observed for ticarcillin, cefuroxime, cefoxitin, cefotaxime, and ceftazidime. The corresponding values for amoxicillin-clavulanate, imipenem, and cefepime were 94.0, 90.0, and 83.0%, respectively.

View this table: [in this window] [in a new window]   TABLE 6.   Results of comparison between Wider system susceptibility testing results and reference broth microdilution MIC for isolates of the family Enterobacteriaceae with known -lactam resistance mechanismsa

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Simultaneous identification and susceptibility testing is clearly an advantage of automatic and semiautomatic commercial systems (4, 7). The Wider system is a newly developed semiautomatic device for bacterial identification and susceptibility testing which needs the classical overnight period for retrieval of results. The major attraction of this system is the image-processing technology that has been applied for the first time to the reading of commercial trays for bacterial identification and susceptibility testing. Biochemical patterns and MICs are processed by the computer and are offered to the operator both visually and, if required, on paper. Indeed, a particular advantage of image-assisted analysis is that the system greatly facilitates reading of the results, but the results for the panels can still be interpreted directly by the microbiologist by applying classical criteria. This eliminates dependence on a fully automatic device; moreover, the technologist can adjust the video-assisted readings prior to the release of a patient's results. This approach has previously been applied for disk diffusion antibiograms (3, 10) and also for an unsuccessful API Aladin instrument, marketed in the late 1980s, which used microdilution tray wells for bacterial identification and susceptibility testing (12).

In the present study, the Wider system correctly identified at the species level 95.7% of routine clinical isolates, represented by 20 genera commonly isolated in a clinical microbiology laboratory. This percentage is similar to or slightly higher than those observed in other studies with other systems (25, 31) and demonstrates that with currently accepted criteria the Wider system is an acceptable method for bacterial identification (15). For the isolates of the family Enterobacteriaceae, the only identification problem was an atypical lactose-positive S. enterica serovar Enteritidis isolate which was misidentified by the Wider system as S. enterica subsp. arizonae. This particular problem has also been delineated with other systems (18); however, the Wider system correctly identified the S. enterica subsp. arizonae ATCC 12323 strain. As with other systems, less accuracy (92.0%) than that observed with isolates of the family Enterobacteriaceae was shown when only NFGNRs were taken into account (8, 22, 24, 25, 27). In all cases, the API system was used as the "gold standard" for bacterial identification, as it is a well-recognized system for this purpose (25, 31). Nevertheless, conventional biochemical tests were also performed when discrepancies between the Wider system and the API galleries were observed before assuming that the API identification was correct.

The Wider system database contains information on 112 and 42 different taxa for gram-negative and gram-positive bacteria, respectively, which includes both organisms usually encountered in clinical laboratories and those rarely isolated. It is remarkable that, with the exception of the oxidase test for gram-negative organisms and catalase and hemolysis tests for gram-positive organisms, the Wider system does not require additional external reactions to provide the final identification or to improve the identification level. Other identification systems need external biochemical tests other than those provided in the microdilution panels to resolve identifications with low probabilities of accuracy (8, 21). As with other devices, if a biochemical profile does not match one in the database, there may be no identification, but this did not occur with the routine isolates tested with the Wider system. As the purpose of our study was to determine the accuracy of the Wider system for the identification of bacteria that would routinely be encountered in a hospital microbiology laboratory, additional studies with other relevant and uncommon organisms should be performed.

The susceptibility testing results obtained with the Wider system in tests with routine clinical isolates showed an overall essential agreement of 95.6% and an overall category interpretation error of 4.2%. Combined major and very major errors were less than 3%. These results meet the performance criteria for susceptibility testing (7, 9, 18). It is of note that these results were compiled by considering the results for both gram-negative and -positive isolates and a large number of antimicrobial agents for each isolate. When analyzed with respect to the organisms tested, the highest essential agreements were observed with isolates of the family Enterobacteriaceae (96.6%) and gram-positive organisms (95.6%). As other investigators have noted with other systems (22, 23), lower essential agreement was observed with NFGNRs (88.0%). Nevertheless, in contrast to automated instrument systems with short incubation times (14, 15), susceptibility testing results could be obtained for all organisms considered to be slowly growing organisms.

When analyzed with respect to the antimicrobial agents tested, the lowest essential agreement for isolates of the family Enterobacteriaceae was observed with imipenem (84.0%). It is remarkable that 26.2% of the imipenem MICs obtained with the Wider system were higher than those obtained by the reference method, but the percentage of interpretive errors was limited to only 1.4 and 0.7% minor and major errors, respectively, with no very major errors. This difference is clearly related to the intrinsic activity of imipenem (modal MIC, 0.25 µg/ml for isolates of the family Enterobacteriaceae) and the relatively high MIC breakpoint for susceptibility (4 µg/ml). For NFGNRs, the essential agreement for imipenem was 88.0%. Again, nearly 25% of the imipenem MICs obtained with the Wider system were higher than those obtained by the reference method. These higher imipenem MICs could be due to insufficient adjustment of inoculum size or to a decline in antimicrobial activity during storage. The former possibility was well corroborated by Doern et al. (6), who demonstrated that an inappropriately large inoculum size in automatic susceptibility testing devices significantly modified the results, particularly for cell wall-active antibiotics. False positive results for resistance may occur when the inoculum size is too large, and the numbers of major and minor errors are increased. In our study, the inoculum for the Wider system was prepared with the Prompt inoculation system (28), as recommended by the manufacturer. In contrast, the inoculum for the reference microdilution method was controlled nephelometrically. On the other hand, it has been shown that the activity of imipenem is particularly affected during storage (5, 26) and is a well-recognized cause of false-positive resistance with commercial microdilution panels (23, 28). It is not a surprise that the imipenem MICs for E. coli ATCC 25922 and P. aeruginosa ATCC 27853 were persistently near the upper limit of the MIC range recommended by NCCLS (20). The same problem of stability during storage could be responsible for the higher penicillin MICs for S. aureus ATCC 29213 and routine gram-positive isolates.

In previous evaluations of susceptibility testing devices, specific attention has been given to resistant isolates (31). In our evaluation, although a limited number of resistant isolates were included among the clinical isolates (data not shown), no problems with specific issues of resistance specifically studied with other automatic susceptibility testing devices were detected, such as oxacillin resistance in staphylococci (13) and high-level aminoglycoside resistance in enterococci (30). Moreover, when assessing susceptibility testing instrument performance, in addition to fresh or stock clinical isolates and quality control strains, a set of organisms with known resistance mechanisms should be included (7, 9). One hundred isolates of the family Enterobacteriaceae with known -lactam resistance mechanisms were studied, and 800 organism-antimicrobial agent combinations were evaluated. Essential agreement and interpretive errors for eight -lactam antibiotics were 94.8 and 5.4%, respectively. Interestingly, the proportion of very major errors with this set of strains was limited to 0.8%. The lowest essential agreement was observed with extended-spectrum and chromosomal class A -lactamase-producing isolates. Slight variations in the inoculum size could affect the amount of these -lactamases and would be the reason for the lower essential agreement. The case of cefepime is illustrative, as this antibiotic has been shown to be very stable during storage (26) but is particularly affected by an increase in inoculum size for those isolates with extended-spectrum -lactamases (R. Cantón and A. Oliver, unpublished data). Interestingly, no decreases in the MICs with the Wider system were observed with ceftazidime and cefotaxime for extended-spectrum -lactamase producing isolates of the family Enterobacteriaceae, thus avoiding false-positive detection of isolates with these enzymes. The design of panels is essential for retrieval of results that can serve as indicators of the presence of extended-spectrum -lactamase-producing isolates (16). The wide range of concentrations for ceftazidime (0.5 to 16 µg/ml) and cefotaxime (0.12 to 8 µg/ml) in the Wider system panels facilitates the detection of these isolates. In addition, the Wider system panels for gram-negative isolates possess the combination of ceftazidime plus clavulanate to facilitate the detection of these -lactamase-producing isolates.

In conclusion, our evaluation showed accurate and acceptable results with both routine clinical isolates and a set of isolates of the family Enterobacteriaceae with known -lactam resistance mechanisms. The Wider system is a new reliable tool which applies the image-processing technology for the reading of commercial trays for bacterial identification and susceptibility testing.