INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Clinical microbiologists and physicians generally agree that it is important for the management of infections caused by gram-negative rods to rapidly and correctly identify these bacteria. For nearly three decades, automated identification systems for gram-negative rods (and other bacteria as well) have been developed and commercialized, but only a few of them (e.g., ATB [bioMérieux], MicroScan [Dade], and VITEK [bioMérieux]) are nowadays significantly present on the market. The new VITEK 2 system (bioMérieux, Marcy l'Etoile, France) differs fundamentally from the previous VITEK system by providing definitive identification results for gram-negative rods (including both members of the family Enterobacteriaceae and nonenteric bacilli) within 3 h (4). This is due to a new fluorescence-based technology that is more sensitive in detecting metabolic changes and that, therefore, by additional continuous monitoring of reactions, provides much faster identifications (4). This paper reports on the evaluation of the new VITEK 2 system for identification of gram-negative rods. The emphasis of our study was on a stress test rather than on a weighted laboratory profile (6). It is concluded that the new VITEK 2 system is a promising new tool for identifying gram-negative rods regarding both speed and accuracy.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Strains, culture conditions, and inoculum preparation. Of the 845 strains included in the present study, 298 came from the culture collections of the Department of Medical Microbiology, University of Zürich, Switzerland, and the Laboratoire de Bactériologie, Faculté de Médecine René-Laennec, Lyon, France. All strains had been identified by established methods (2, 3, 7, 11, 13), and the identities of a few of them had also been confirmed by the Nosocomial Pathogens Laboratory Branch at the Centers for Disease Control and Prevention (Atlanta, Ga.) as well as the Special Bacteriology Laboratory at the same institution. The other 547 strains included were fresh clinical strains (<4 weeks old) isolated in either of the two laboratories contributing to this study. These strains had also been identified by established methods (2, 3, 7, 11, 13). Discrepancies that occurred between the laboratory identification and the VITEK 2 identification were resolved with the API 50 CHE, ID 32 GN, and API 20 NE systems, as well as the biotype 100 carbon substrate assimilation panel (all from bioMérieux).

ID-GNB card and VITEK 2 system. The identification card for gram-negative bacilli (ID-GNB card) for the VITEK 2 system is a 64-well plastic card containing 41 fluorescent biochemical tests, including 18 enzymatic tests for aminopeptidases and -osidases. Substrates used for detection of aminopeptidases are usually coupled with 7-amino-methylcoumarin (7AMC); substrates for detection of -osidases are usually coupled with 4-methylumbelliferone (4MU). The 18 test substrates are as follows: 4MU--arabinopyranoside, 4MU--D-galactoside, -L-glutamic acid-7AMC, 4MU--D-cellobiopyranoside, 4MU--D-galactoside, 4MU--D-glucoside, 4MU--D-glucuronide, 4MU--D-mannopyranoside, 4MU-N-acetyl--D-glucosaminide, 4MU-N-acetyl--D-galactosaminide, 4MU--D-xyloside, glutaryl-glycyl-arginine-7AMC, -L-glutamic acid-7AMC, 4MU-phosphate, L-proline-7AMC, L-pyroglutamic acid-7AMC, L-lysine-7AMC, and Z-arginine-7AMC. Furthermore, the ID-GNB card includes 18 fermentation tests (adonitol, L-arabinose, D-cellobiose, D-galacturonate, D-glucose, glucose-1-phosphate, D-glucuronate, inositol, 5-keto-gluconate, D-maltose, D-mannitol, D-melibiose, palatinose, D-raffinose, L-rhamnose, sucrose, D-sorbitol, and D-trehalose), 2 decarboxylase tests (ornithine and lysine), and 3 miscellaneous tests (urease, utilization of malonate, and tryptophane deaminase).

Quality control. Eight strains were used as quality controls every day during the evaluation. The strains were Brevundimonas diminuta ATCC 11568, Enterobacter sakazakii ATCC 51329, Klebsiella pneumoniae ATCC 35657, Klebsiella oxytoca ATCC 43863, Proteus vulgaris ATCC 13315, Shigella sonnei ATCC 25931, Sphingobacterium spiritivorum ATCC 33861, and Stenotrophomonas maltophilia ATCC 17666. All eight quality control strains had to be identified correctly in order to allow identification of the test strains.

Reporting of results. Identification scores provided by the VITEK 2 software were not considered; rather, the interpretation of the results given by the software was used. There were five different categories of results: (i) "correctly identified" meant that a strain was unambiguously correctly identified at the species level (i.e., the correct identification was the only one given); (ii) "low discrimination resolved" meant that the correct identification was obtained after simple, immediate additional tests (oxidase, hemolysis, indole, motility, and pigmentation) were performed; (iii) "low discrimination not resolved" meant that the correct identification was given, among others, but that the simple, immediate additional tests did not lead to the final correct identification of the strains; (iv) "misidentified" meant incorrectly identified strains; (v) "not identified" meant that no identification was given at all.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

The results of the testing of 845 strains are listed in Table 1. Six hundred and fifty (76.9%) strains were Enterobacteriaceae, and 195 (23.1%) were nonenteric bacilli. Overall, 84.7% of all bacteria were correctly identified (88.2% of the Enterobacteriaceae and 73.3% of the nonenteric bacilli). Thirty-two (3.8%) strains were correctly identified by additional, simple, rapid tests (see Reporting of results, above). About half of the 80 (9.5%) strains not identified at the species level after application of the five simple additional tests belonged to the Enterobacteriaceae (42 of 80), and the other half belonged to the nonenteric rods (38 of 80). Only 7 strains (0.8%) were misidentified, and 10 (1.2%) were not identified (which implied not that the organisms were misidentified but that they were treated as such).

No significant differences were observed between the two laboratories that tested the system: one institution found 86.7% correct identifications, 2.2% identifications of "low discrimination resolved", 9.1% identifications of "low discrimination not resolved", 0.8% misidentifications, and 1.2% no identifications, whereas the other laboratory observed 82.2%, 5.9%, 10.0%, 0.8%, and 1.1%, respectively.

To carry out a stress test of the system, we included strains belonging to 70 different taxa in our evaluation (Table 1). However, more strains of the most frequently encountered gram-negative bacilli in the routine clinical laboratory, namely Escherichia coli (8.6% of all isolates tested), Pseudomonas aeruginosa (5.4%), Salmonella spp. (5.1%), and K. pneumoniae subsp. pneumoniae (5.0%), were tested than other and more rarely isolated species. When combined, 96.1% of the strains belonging to these four taxa were correctly identified.

In a pragmatic approach, the manufacturer had categorized infrequently encountered and relatively inert nonfermenting bacilli in a group designated "various nonfermenting gram-negative bacilli" (Table 1). These were responsible for 31 of 80 (38.8%) cases in which the identification was interpreted as "low discrimination not resolved". If these strains were excluded from the evaluation, 88.0% of the strains would have been correctly identified; an additional 3.9% would have been correctly identified after application of simple additional tests. Other taxa with a relatively high percentage of "low discrimination not resolved" results included Enterobacter intermedius, Klebsiella planticola, and Klebsiella terrigena. However, the system evaluated did not claim to identify the latter two species (which are found very rarely in humans) but these two species always appeared together with Klebsiella oxytoca and K. pneumoniae subsp. pneumoniae as identification.

The reliability and reproducibility of the system were demonstrated by the fact that during only 3 days of the entire evaluation period of 1.5 months was one of the eight quality control strains not correctly identified. The stability of the system was also demonstrated by retesting the 7 misidentified strains and the 10 nonidentified strains for which the same results were observed upon retesting.

The strains which were misidentified or not identified did not belong to any particular taxon but were distributed over different taxa (Table 1). The problematic reactions for the misidentified strains are outlined in Table 2. None of these reactions was significantly more frequently observed than others.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The evaluation presented in this report was mainly a stress test of the system since its database was challenged by a diverse group of organisms, including species which are very rarely encountered in the routine clinical laboratory. An accuracy rate of 88.5% identification to the species level after 3 h (including the strains for which simple, rapid tests had to be performed) is, in our view, acceptable, although other authors demand a 90% accuracy level (8). The critical point is how to interpret the data for the 9.5% of strains which were identified as "low discrimination not resolved." The VITEK 2 database does not recommend further tests other than those five simple, immediate ones because the system is aiming at rapid identification for which time-consuming supplementary tests are contraindicated and/or not often performed in a routine clinical laboratory. However, experienced clinical microbiologists may easily find and carry out additional tests which may eventually lead to the identification at the species level of the 9.5% of strains identified as "low discrimination not resolved."

It is likely that if a system performs well in a stress test (like the VITEK 2 system in conjunction with the ID-GNB card) it will also do so in a weighted laboratory profile (6). Therefore, it is predicted that the evaluated system may also perform well under the conditions of a routine clinical laboratory.

The fact that nonenteric bacilli were not identified as well as Enterobacteriaceae can be explained by the slower metabolism of some nonenteric bacteria, leading to ambiguous results in the reaction wells. It has also been observed in evaluations of other automated identification systems for gram-negative bacteria that nonenteric bacilli are usually not identified as well as Enterobacteriaceae (8, 9, 10, 12).

One major advantage of the new VITEK 2 system is its speed in reliably identifying gram-negative rods within 3 h. This is basically achieved by the more sensitive fluorescence-based technology used in the system. By increasing the number of substrates from the previous Vitek GNI+ (30 tests) to the ID-GNB card (41 tests), a broader and more detailed database has been built by the company and allows a better discrimination between related taxa. However, even the more sensitive fluorescence-based technology used in the ID-GNB card did not significantly change the outcome of the identifications of some slowly metabolizing nonfermenting bacteria, which were categorized as "various nonfermenting gram-negative bacilli." Additional taxa included in this group which were not explicitly tested by us included Alcaligenes spp., Bordetella avium, Bordetella bronchiseptica, CDC group IVc-2 bacteria, Comamonas spp., Pseudomonas alcaligenes, Pseudomonas mendocina, Pseudomonas pseudoalcaligenes, and Oligella spp. Scientifically, it might be desirable to identify every strain (even the nonfermenters) at the species level, but it has been stated before, and we agree with this opinion, that identification of certain members of non-Enterobacteriaceae to the species level may be unnecessary (6), particularly from the clinical point of view. Other medically relevant gram-negative rods which were not tested in our evaluation include Agrobacterium radiobacter, Chryseobacterium meningosepticum, Flavimonas oryzihabitans, Brucella spp., and Burkholderia pseudomallei. It is important to realize that the conclusions drawn in this paper apply to the tested taxa only and that the performance of the VITEK 2 system for some rarely encountered nonfermenting gram-negative rods is not known at present and requires further investigations.

Obviously, the VITEK 2 system in conjunction with the ID-GNB card represents an improvement regarding speed compared with the previous VITEK system. In our evaluation, 88.5% of all strains were correctly identified after 3 h, whereas in the evaluation of O'Hara et al., applying the previous GNI+ card, only 47% of all enteric strains were identified in 3 h or less (8). Robinson et al., applying the previous GNI+ card, observed a cumulative percentage of 50.8% correct identifications after 4 h when a less diverse group of organisms was tested, which included 20.9% E. coli strains and 15.6% P. aeruginosa strains (10). Pfaller et al., also using the GNI card on the VITEK system, found after 4 h 58% of the enteric bacteria and 15% of the nonenteric bacteria correctly identified (9).

As for the work flow in a routine clinical bacteriology laboratory, the VITEK 2 system could be integrated like any other automated identification system. It seems to be possible to start the identification of gram-negative rods from primary cultures, since, for inoculation of a ID-GNB card, a suspension of a 0.5 McFarland standard only has to be prepared in a small volume of saline. Other advantages of the VITEK 2 system are the decreased turnaround and hands-on times since the system is nearly fully automated. We calculated a hands-on time of about 20 min for 10 strains, which included collection of all needed material, preparation of the suspension, filling and loading of the cards into the system, and collection of the computer printouts or review of the identifications. The high degree of automation may also improve accuracy. It has been demonstrated previously that the results obtained with the ID-GNB card are independent of the media (except eosine methylene blue agar and salmonella-shigella agar) on which strains are cultured (1). Factors affecting the quality of the identification are the age of the culture (8- to 24-h cultures are best) and the inoculum (McFarland standard of 0.5 or slightly higher is best) but not the age of the inoculum suspension (5). During the evaluation we did not encounter difficulties regarding air bubble formation while filling the cards. Gas formation within the reaction wells by some species during the kinetic reading was taken into account by a specific software algorithm. The ID-GNB card can be considered safe and resistant to contamination as it is a fully closed system to which no reagents have to be added.

As mentioned before, the database contains a larger number of taxa than are usually encountered in the ordinary routine clinical laboratory. It was noted that the taxonomy used in the database was up to date. Furthermore, it is emphasized that our evaluation data were valid at the time of our evaluation but that another important value of a commercial identification system lies beyond this performance and must be the capability of manufacturers to maintain or even improve the performance of an identification system over time.

Unfortunately, the effective costs of performing an analysis on the VITEK 2 using the ID-GNB card could not be calculated at the time of writing this article since the system has not been introduced into the market. However, costs for labor are minimized since the number of manual steps needed has been reduced to a minimum.

Evaluations of the VITEK 2 system (applying the ID-GNB card) by other authors, in particular by directly comparing it with other manual or automated identification systems, are encouraged in order to confirm or contradict the results of our study.