Performances of VITEK 2 Colorimetric Cards for Identification of Gram-Positive and Gram-Negative Bacteria

The purpose of this study was to evaluate the new VITEK 2 identification cardsthat use colorimetric reading to identify gram-positive and gram-negativebacteria (GP and GN cards, respectively) in comparison to fluorimetriccards (ID-GPC and ID-GNB, respectively). A total of 580 clinicalisolates and stock collection strains belonging to 116 taxa wereincluded in the study. Of the 249 gram-positive strains tested withboth the ID-GPC and GP cards, 218 (87.5%) and 235 (94.4%) strains werecorrectly identified (to the genus and species level), respectively.Of the 331 gram-negative strains tested with the ID-GNB andGN cards, 295 (89.1%) and 321 (97%) strains were correctly identified,respectively. Another focus of the study was to apply the percentagesof correct identifications obtained in this study to the listof bacteria isolated in our laboratory (32,739 isolates) inthe year 2004. We obtained 97.9% correct identifications withthe colorimetric cards and 93.9% with fluorescent cards.

INTRODUCTION

Top

Introduction

The correct and rapid identification of gram-negative and gram-positive bacteriain clinical microbiology is the first step in the interpretationof antimicrobial susceptibility tests for correct treatmentof patients (8). Although the VITEK 2 system (bioMérieux,Marcy l'Etoile, France) combined with the ID-GNB and ID-GPCcards allowed an identification within 3 h using fluorescencereading, the weakness of this system was the breadth of itsidentification database, especially for nonfermenting bacilli,such as Pseudomonas spp. and Acinetobacter, and for gram-positivecocci, such as Streptococcaceae (2, 5). New cards (GP and GN cards)will soon be available that use colorimetric reading. Thesecards are suited to the VITEK 2 system, improving the identificationof nonfermenting bacteria and gram-positive cocci (3, 4). Theaims of this study were (i) to evaluate the performances ofthe new colorimetric cards in comparison to fluorimetric cards(ID-GNB, ID-GPC) to identify 580 clinical isolates and stockcollection strains belonging to 116 taxa and (ii) to determinethe accuracy obtained with both readings by applying the percentagesof correct identifications obtained in this study with the colorimetricand fluorescent cards to the list of bacteria isolated in ourlaboratory in the year 2004.

MATERIALS AND METHODS

Top

Materials and Methods

Strains. A total of 580 strains were tested, consisting of 331 gram-negativebacilli and 249 gram-positive cocci belonging to 68 and 48 taxa,respectively. Clinical isolates were collected over 6 monthsfrom nonconsecutive patient cultures and selected either toobtain around 20 strains of the most frequently isolated speciesor to be in agreement with the distribution of isolates annuallyrecovered in the laboratory. In order to have an idea of theperformance of testing for the most rarely isolated species,a panel of 181 microorganisms was selected from the laboratorystock collection.

Identification protocol. All isolates were cultured onto Columbia agar with 5% horseblood (18 to 24 h at 35°C) to ensure purity and viability.Stock strains were subcultured twice. Microorganisms were testedseparately with two VITEK 2 instruments: the first for fluorescencereading, the second for colorimetric reading. The new VITEK2 cards and the upgraded VITEK 2 were previously described (3).Both systems were used according to the recommendations of themanufacturer. Bacterial suspensions were made in 0.45% sodiumchloride solution and adjusted to a McFarland standard of 0.50to 0.63 by using a Densicheck system (bioMérieux). Identicalinocula of each strain were tested in parallel using fluorimetriccards (ID-GNB, ID-GPC) and colorimetric cards (GN, GP) accordingto the manufacturer's instructions. When the identificationresults were different between the fluorimetric and the colorimetriccards, the strain was retested with the both methods. In caseof persistent discrepancy, the strain was identified with API strips(ID-32 Staph, Rapid ID-32 Strept, ID-32 E, ID-32 GN, API 20NE; bioMérieux)to resolve the identity of the strain. When there was a mismatchbetween identifications obtained with both the VITEK 2 cards andthe API strips, isolate identification was determined by DNA sequencingof the 16S rRNA (Microseq 500; Applera, Foster City, Calif.) (1, 10)and/or sodA (9), and/or rpoB (7) gene.

Quality controls. Fifteen strains were used as quality controls every 2 monthsduring the evaluation. For gram-positive bacteria, the strainswere Staphylococcus saprophyticus ATCC BAA 750, S. aureus subsp.aureus ATCC 29213, Kocuria kristinae ATCC BAA 752, Listeria monocytogenesATCC BAA 751, Streptococcus thermophilus ATCC 19258 T, S. sciuriATCC 29061, Enterococcus casseliflavus ATCC 700327, and S. equisubsp. zooepidemicus ATCC 43079. For gram-negative quality control, thestrains were Klebsiella oxytoca ATCC 700324, Acinetobacter baumanniiBAA 747, Enterobacter cloacae ATCC 700323, Ochrobactrum anthropiATCC BAA 749, Proteus vulgaris ATCC 6380, Shigella sonnei ATCC25931, and Stenotrophomonas maltophilia ATCC 17666. Each qualitycontrol strain was tested with both the VITEK 2 fluorimetricsystem and the VITEK 2 colorimetric system.

Data analysis. Results were separated into four groups: first-choice identification (i.e.,the determination of the genus-and-species level was the same withboth systems and was given as excellent, very good, good, or acceptable);low discrimination (the determination resulted in a choice amongtwo or four species with different values of T indexing needinga few supplemental procedures such as oxidase, motility, indole, pigmentation,or hemolysis testing to determine the correct identification);misidentification (the determination resulted in incorrect identification);and no identification (the determination resulted in doubtful,unacceptable, or unreliable identification). Correct identificationwas defined as the association of first-choice identificationand low discrimination. Results were expressed in numbers andpercentages.

Two supplemental analyses of the data were carried out. Thefirst analysis was performed to determine the levels of accuracyobtained with the two systems by comparing the bacterial speciesidentified in the year 2004 in our laboratory and isolated fivetimes or more to the bacterial species tested in the study.The second analysis was to perform the same evaluation withthe 17 gram-negative rods most frequently recovered from bloodcultures in the microbiology laboratories of 33 French universityhospitals (11).

RESULTS AND DISCUSSION

Top

Results and Discussion

Of the 249 gram-positive strains tested with both the ID-GPCand GP cards and the 331 gram-negative strains tested with boththe ID-GNB and GN cards, 218 (87.5%), 235 (94.4%), 295 (89.1%), and321 (97%) were correctly identified (to the genus or species level),respectively (Tables 1 and 2). A total of 33 bacteria remainedunidentified with the fluorimetric cards, whereas the colorimetriccards did not give an identification for five strains. Regardlessof their origin (combined stock collection and clinical isolates),fermenting gram-negative bacteria were correctly identifiedwith both ID-GN and GN cards (97.2% and 98.7%, respectively).In contrast, nonfermenting gram-negative bacteria were betteridentified with GN cards (92.1%) than with ID-GN cards (67%). Gram-positivebacteria were better identified with colorimetric cards thanwith fluorimetric cards. For Streptococcaceae, readings of fluorimetricand colorimetric cards gave 87.9% and 90.9% correct identifications,respectively. For Micrococcaceae, these readings were 87.2%and 98.3%, respectively. Our results obtained with gram-negativebacteria were for the most part in agreement with those reportedby Funke and Funke-Kissling (3). Testing 511 fermenting and144 nonfermenting gram-negative bacilli, these authors (3) obtainedslightly better results with the GN cards than we did in thepresent study (99.5% and 98.7%, respectively). Recently, inanother study focusing on gram-positive bacteria, Funke andFunke-Kissling (4) obtained correct identification of Streptococcaceae(217 strains) and Micrococcaceae (147 strains) for 99.1% and99.3% of these bacteria, respectively, whereas the results were90.9% and 98.3% in our study. It should be noticed that theresults reported by Funke and Funke-Kissling (3, 4) appearedbetter but that the numbers of taxa tested (13, 18, and 12 for Micrococcaceae,Streptococcaceae, and nonfermenting bacilli, respectively) werelower than in our study, except for fermenting bacteria (42taxa in both studies). In fact, as Funke and Funke-Kisslinghave previously claimed to have done (3, 4), we have selectedrare bacteria isolated in routine practice and some of them,such as S. constellatus or S. gordonii, while infrequently isolated inroutine testing, were isolated in large numbers in our study(for example, eight strains of S. constellatus instead of onein Funke's study).

View this table:
[in this window]
[in a new window]
TABLE 1. Identification of gram-positive bacteria by fluorimetric and colorimetric VITEK 2 cards

View this table:
[in this window]
[in a new window]
TABLE 2. Identification of gram-negative bacteria by fluorimetric and colorimetric VITEK 2 cards

The GP and GN identification cards contain new tests (16 and21 tests for GP and GN, respectively) allowing an improvementof the VITEK 2 databases; in fact, 57 and 38 new species ofgram-positive cocci and gram-negative bacilli were added tothe database. Of these, seven gram-positive species (eight strains)were tested and only one strain of Pediococcus pentosaceus wasnot correctly identified. Of the 17 new gram-negative speciestested (36 strains), only four species (four strains) were notcorrectly identified: Bordetella bronchiseptica, Pseudomonasalcaligenes, and P. putida were misidentified (one strain each),and one strain of P. fluorescens was not identified. Thus, themisidentification percentages obtained with the colorimetriccards ranged from 2.1% to 4.8%, results which were slightlybetter than those obtained with the fluorescent cards (5.1%to 6.8%). These misidentified bacteria were observed only withStreptococcus spp. such as S. constellatus (three strains) orS. gordonii (two strains) or nonfermenting gram-negative bacillisuch as Acinetobacter spp. belonging to genomospecies 1, 2,3, or 13 (A. calcoaceticus/A. baumannii complex). Compared toprevious study results (2, 5, 6), the database was enlargedand improved, especially for nonfermenting bacilli and Micrococcaceae.The database was also enlarged to include some gram-positivebacilli such as Erysipelothrix rhusiopathiae and six speciesof the Listeria genus.

The second aim of this study was to evaluate the performanceof the new colorimetric cards in routine practice. Thus, weapplied the percentages of correct identifications obtainedin this study with the colorimetric and fluorescent cards tothe list of bacteria isolated in our laboratory in the year2004. From the species included in the database of the VITEK 2,71 species were selected, representing 32,739 bacteria isolatedfive times or more in 2004. An overall correlation of 97.9%correct identifications for gram-positive and gram-negativebacteria was obtained, whereas it was equal to 93.9% with fluorescentcards. The same determination was performed with a selectionof 17 gram-negative taxa isolated more frequently in 33 Frenchuniversity hospitals (11). Identification with colorimetriccards gave an overall identification to the species level of99.7%, whereas it was equal to 95.9% with fluorescent cards.

The VITEK 2 system, equipped for colorimetric reading of thenew GP and GN cards, keeps the advantages of the VITEK 2 (2, 5, 6, 8),i.e., reliable identification, fully automated incubation andinterpretation, and minimal supplemental testing required. Theresults provided by the colorimetric VITEK 2 may be consideredaccurate due to the improvement and the extension of its database,mainly for nonfermenting bacteria and Streptococcaceae. In thisstudy, the identifications of bacteria were provided between5.2 h (fermenting bacteria) and 6.7 h (nonfermenting bacteria), whichwas slightly greater than the time required for reading with fluorescentcards. However, the results were always provided within a day.In conclusion, the results obtained in this study demonstratethe good performances of the new VITEK 2 cards, allowing theiruse in routine practice with a highly acceptable level of identification accuracy.

ACKNOWLEDGMENTS

We thank bioMérieux for kindly providing the VITEK 2system and Geneviève Bossy and Marie-Christine Saccomanifor their technical assistance.

FOOTNOTES

* Corresponding author. Mailing address: Laboratoire de Bactériologie-Hygiène, Hôpital A. Calmette, F-59037 Lille Cedex, France. Phone: 33 320 444 908. Fax: 33 320 444 895. E-mail: rcourcol{at}chru-lille.fr.

REFERENCES

Top