Comparative evaluation of the VITEK 2 Advanced Expert System (AES) in five UK hospitals

Transcription

1 Journal of Antimicrobial Chemotherapy (2003) 51, DOI: /jac/dkg234 Advance Access publication 14 April 2003 Comparative evaluation of the VITEK 2 Advanced Expert System (AES) in five UK hospitals J. Barry 1, A. Brown 2, V. Ensor 1, U. Lakhani 3, D. Petts 4, C. Warren 3 and T. Winstanley 5 * 1 Department of Microbiology, University Hospital, Birmingham; 2 Department of Microbiology, Ninewells Hospital, Dundee; 3 Department of Infection, Guy s & St Thomas Hospital, London; 4 Department of Microbiology, Basildon Hospital, Basildon, Essex; 5 Department of Microbiology, Royal Hallamshire Hospital, Glossop Road, Sheffield, UK Received 3 January 2003; returned 18 February 2003; revised 18 February 2003; accepted 26 February 2003 Objectives: We carried out an evaluation of VITEK 2 in five UK laboratories, comparing results with gold standard agar-dilution MIC data, assessing its ability to recognize resistant phenotypes and comparing results with those generated by routine antimicrobial susceptibility testing methods. Methods: Laboratories tested a collection of 82 strains selected on the basis of their challenging and characterized resistance mechanisms. Results: In comparison with the reference MIC method, VITEK 2 gave an essential agreement of 304/315 (enterococci), 1619/1674 (staphylococci) and 2937/3074 (Gram-negative bacilli): overall 96.0% agreement. Corresponding category (SIR) agreements with VITEK 2 were 247/252, 1496/ 1561 and 2478/2626 (overall 95.1%). Using five routine methodologies, category agreements ranged from 58/63 to 45/45; 222/232 to 174/174, and 333/372 to 250/259 for the three organism groups with an overall agreement of 95.0%. In contrast to VITEK 2 Advanced Expert System (AES), routine microbiology laboratories did not attempt to detect resistance mechanisms for every antibiotic studied. VITEK 2 AES detected all 19 resistance mechanisms in enterococci: where applicable, routine methods detected 14, 10 and 10. Of 30 resistance mechanisms in staphylococci, VITEK 2 AES detected 25 compared with 23, 20, 17 and 18 detected by routine methods. Finally, of 44 resistance mechanisms in Gram-negative bacilli, VITEK 2 AES detected 30 compared with 30, 23, 15 and 10 detected by routine methods. Conclusions: VITEK 2 performed susceptibility tests accurately and the AES detected and interpreted resistance mechanisms appropriately. Heavy inocula in a liquid medium possibly favour better expression of certain resistance determinants. Although certain routine microbiology methods performed adequately, VITEK 2 AES offers a rapid, standardized method suited to laboratories lacking experience of resistance mechanisms and/or those not testing an appropriate number, or range, of antibiotics to detect resistance phenotypes. Keywords: VITEK 2; Expert; automated AST Introduction VITEK 2 (biomérieux, Marcy l Étoile, France) 1 3 is an integrated system that automatically performs rapid identification using algorithms based on fluorescence and colorimetry and antimicrobial susceptibility testing (AST) based on kinetic analysis of growth data. The Advanced Expert System (AES) analyses the AST data using a knowledge base of 2000 phenotypes and MIC distributions. Interpretative reading of susceptibility tests aims to analyse susceptibility patterns rather than results for individual antibiotics so that underlying resistance mechanisms may be predicted. It... *Corresponding author. Tel: ; Fax: ; tgwinstanley@hotmail.com The British Society for Antimicrobial Chemotherapy

2 J. Barry et al. allows anomalous combinations of phenotype and organism to be reconsidered; prediction of further antibiotics worth testing; suppression of tenuous results, and surveillance of the prevalence of resistance mechanisms. 4 6 VITEK 2 and AES have been evaluated in several countries Here, we have carried out an evaluation in five UK laboratories and, in addition to comparing VITEK 2 MIC data with reference MIC data and assessing the ability of VITEK 2 AES to recognize resistance phenotypes, we have extended our study to compare VITEK 2 with routine antimicrobial susceptibility testing methods. Materials and methods Strains and evaluation strategy VITEK 2 systems were installed and verified in five UK laboratories by biomérieux Application Specialists. Each laboratory was asked to test the same set of 82 strains (Enterobacteriaceae 36; pseudomonads 8; Staphylococcus aureus 22; coagulase-negative staphylococci 7; enterococci 9) obtained from the Institut Pasteur collection and distributed by biomérieux; none of these strains had been used for the initial development of VITEK 2 AES. Identification of antibiotic resistance mechanisms was achieved using appropriate biochemical and/or molecular procedures, including determination of the isoelectric focusing substrate profile, determination of the inhibitor profile, recombinant DNA techniques, DNA amplification and sequencing. 9 The evaluators were informed of the species, but not of the genotypes or anticipated resistance phenotypes. In total, the 82 strains displayed 95 phenotypes (Table 1). MICs were determined, by Professor Courcol (University Hospital of Lille, France), on nine separate occasions using an agar dilution method according to NCCLS guidelines 12 and these data were used as gold standard reference values for data comparison. Susceptibility tests with VITEK 2 Strains were sub-cultured twice, then grown for h at 35 C on Columbia agar containing 5% horse blood (biomérieux) in air. Suspensions of these cultures were made in 0.45% saline, adjusted to the turbidity of a 0.6 McFarland Standard, and used to load the test cards for VITEK 2, which was used in accordance with the manufacturer s directions. The following test cards were used: Enterobacteriaceae (AST-N020); pseudomonads (AST-N017); staphylococci (AST-P523), and enterococci (AST-P524). Escherichia coli ATCC 25922, Pseudomonas aeruginosa ATCC 27853, S. aureus ATCC and ATCC and Enterococcus faecalis ATCC were used as control strains. Antibiotic resistance phenotypes were determined using VITEK 2 AES. Susceptibility tests with local methodology Centres 1 and 2 both used an agar-incorporation, breakpoint method based on that described by Faiers et al. 13 Both centres used Iso-sensitest agar (Oxoid, Basingstoke, UK) with an inoculum of 10 4 cfu/spot and both read plates using a Mastascan Elite (Mast Group Ltd, Bootle, UK). Centre 1 prepared antibiotic plates using antibiotic powders, of known potency, obtained from the relevant manufacturers and used BSAC interpretive criteria 14,15 throughout. Resistance phenotypes were determined using an Expert system running on Microsoft Access [T. Winstanley & K. Humphries (Mast Group Ltd, Bootle, UK), unpublished data] largely populated with published rules. 4 6 Centre 2 prepared plates using Adatabs (Mast Group Ltd) and also used BSAC interpretive criteria 14,15 with the exception of ampicillin for enterococci (1 mg/l) and ofloxacin for Enterobacteriaceae (1 mg/l). Lysed horse blood (5%) was added to the medium for testing Gram-positive cocci and the production of β-lactamase by staphylococci was detected using β-lactamase test papers (Intralactam; Mast Group Ltd) after induction on methicillin plates (0.5 mg/l). Centre 2 determined antibiotic resistance phenotypes with VITEK 2 only. Centres 3 and 4 followed the BSAC standardized disc diffusion susceptibility method. 14,15 Centre 3 read plates manually but used rule-based computerized checking as well as manual interpretation of phenotypic resistances. Centre 4 used an Aura image automated zone reader (Oxoid, Basingstoke, UK) 16 to read inhibition zones; disc interaction tests to determine the presence of β-lactamases 17,18 and Etests (Cambridge Diagnostic Services, Cambridge, UK) to confirm glycopeptide resistance in enterococci; resistance phenotypes were interpreted manually. Centre 5 used VITEK (biomérieux; version VTK-R07.01) according to the manufacturer s instructions. GNS-532 test cards were used for Enterobacteriaceae and pseudomonads; GPS-519 cards were used for staphylococci and enterococci; NCCLS criteria were used. 12 All Expert rules (XPT-ON) that would routinely be used were enabled. Appropriate control organisms were examined by every method. Retention of key resistances was not confirmed at the time of the study. However, all organisms giving aberrant results in terms of MIC or category of susceptibility, together with a representative sample of those giving concordant results, were re-examined using Etests (Cambridge Diagnostic Services, Cambridge, UK). Data analysis The percentage of essential agreement between the test MIC and the MIC determined by the reference agar-dilution method was defined as the percentage of results within ±1 log 2 dilution MIC. MIC results were categorized (susceptible, intermediate, resistant) using NCCLS 12 or BSAC 14,15 interpretive criteria depending on the methodology of the comparator. All test results were analysed before antibiograms being edited on 1192

3 VITEK 2 Advanced Expert System Table 1. Relevant phenotypes and/or genotypes studied Relevant phenotypes and/or genotypes Total Bacterial species (n) Enterococci Overproduction of PBP5 1 E. faecium (1) Resistance to aminoglycosides 4 high-level GEN [aac(6 )-aph(2 )] E. faecalis (2) high-level KAN, TOB [ant(4 )-Ia] E. faecium (1), E. gallinarum (1) Resistance to glycopeptides 7 VAN, TEC (vana) E. faecium (1) VAN (vanb) E. faecalis (2), E. faecium (2) VAN (vanc) E. gallinarum (2) MLSB (ermb) 4 E. faecalis (2), E. faecium (2) Resistance to tetracyclines 3 E. faecalis (1), E. faecium (1), E. gallinarum (1) Staphylococci Penicillinase 2 S. haemolyticus (2) Methicillin resistance (meca) 13 S. aureus (10), S. epidermidis (2), S. warneri (1) Resistance to KAN, GEN, TOB [aac(6 )-aph(2 )] 1 S. aureus (1) Resistance to MLS antibiotics 3 efflux S. haemolyticus (2) lincosamide nucleotidylation (lina, lina ) S. haemolyticus (1) Resistance to fluoroquinolones 6 gyra, parc; parc; gyra S. aureus (3) efflux S. aureus (1) parc + efflux; gyra + parc + efflux S. aureus (2) Resistance to rifampicin 2 high-level (1) and low-level (1) resistance S. aureus (2) Resistance to teicoplanin 3 S. aureus (3) Gram-negative bacilli Resistance to aminoglycosides 4 GEN [aac(3)-ia] E. coli (1) AMK, KAN, NET, TOB [aac(6 )-Ia] E. coli (1) KAN, GEN, TOB [ant(2 )-Ia] E. coli (1) GEN, NET, TOB [aac(2 )-Ia] P. stuartii (1) Extended-spectrum β-lactamases 10 E. coli (6) (TEM and SHV derivatives) K. pneumoniae (1) C. diversus (1), P. mirabilis (1), P. stuartii (1) (TEM or SHV or PER derivatives) 1 P. aeruginosa (1) High-level cephalosporinase 3 E. cloacae (1), C. freundii (1), C. braakii (1) High-level cephalosporinase + porin alteration 1 E. cloacae (1) Plasmid-mediated cephalosporinase (ampc) 1 K. pneumoniae (1) Cephalosporinase 1 E. coli (1) Acquired penicillinase 8 E. coli (3) P. mirabilis (1) K. pneumoniae (1) P. aeruginosa (3) High-level chromosomal β-lactamase 1 K. oxytoca (1) Inhibitor-resistant penicillinase 2 E. coli (2) Resistance to imipenem 4 impermeability P. aeruginosa (2) imipenemase E. cloacae (1), S. marcescens (1) Resistance to tetracyclines 4 E. coli (4) Resistance to quinolones 6 nalb E. coli (1) nald E. coli (1) gyra E. coli (2) gyra + pare E. coli (2) Total 95 GEN, gentamicin; KAN, kanamycin; TOB, tobramycin; VAN, vancomycin; TEC, teicoplanin; MLSB, macrolides, lincosamides, streptogramins B; AMK, amikacin; NET, netilmicin. 1193

4 J. Barry et al. the basis of inferred resistance mechanisms. The relevant susceptibility breakpoints were used to calculate very major, major and minor errors between the agar dilution MIC and the comparator. Very major errors corresponded to organisms for which the MIC indicated resistance and the test method indicated susceptibility (i.e. total number of resistant strains was used as the denominator). Major errors corresponded to organisms for which the MIC indicated susceptibility and the test method indicated resistance; minor errors corresponded to organisms for which the MIC indicated intermediate resistance and the test method indicated susceptibility or resistance and vice versa. Resistance mechanisms inferred by VITEK 2 AES or the individual centres were compared with previous conclusions based on biochemical or genetic analysis (Table 1). Results were graded as Agreement when the same mechanism was inferred, as Partial agreement when the correct mechanism amongst others was inferred and as Disagreement when a different mechanism was inferred. Results The MICs for the five quality control strains determined in the five centres were all within one doubling dilution of each other and gave results within acceptable ranges for VITEK 2 (data not shown). Excluding these, a total of 5063 MICs were determined with VITEK 2 in the five test centres. Of these, six gave consistently aberrant results using both local methods and VITEK 2 compared with the reference agar-dilution method, and susceptibility categories were amended on the basis of Etest results, used to confirm retention of key resistances. These comprised a strain of Enterococcus gallinarum with vanc, originally considered susceptible to vancomycin; a strain of Enterococcus faecium originally considered to have a penicillin MIC of 2.0 mg/l but found to have an MIC of mg/l on re-testing; a strain of S. aureus with an acquired penicillinase, originally considered susceptible to penicillin and three strains of ESBL-producing Enterobacteriaceae originally considered susceptible to piperacillin. In comparison with the agar-dilution reference method, VITEK 2 gave an overall essential agreement of 96.5%, 96.7% and 95.5% for enterococci, staphylococci and Gramnegative bacilli, respectively (Table 2). Inter-laboratory variations were minimal. Results showed that 3.5% of MICs determined for Gram-negative bacilli with VITEK 2 were greater (>1 dilution) but that only 1.0% were lower (<1 dilution) than those obtained with the reference method. This skew in distribution occurred with ofloxacin, gentamicin, tobramycin, amikacin, piperacillin, cefalothin and ceftazidime but was most noticeable with cefotaxime and with resistant rather than with susceptible organisms. Results for these antibiotics accounted for the lowest degrees of agreement. Oxacillin MIC results with staphylococci tended to be one dilution lower than those obtained with the reference method but, overall, the distribution of MICs with staphylococci was towards higher MICs than with the reference method (2.3% versus 0.9%) and essential agreement with individual antibiotics was good. MICs with enterococci were evenly distributed although essential agreement with nitrofurantoin was only 88.9%: five strains (three of these the same strain tested on different occasions) gave MICs higher than the reference MIC. VITEK 2 results from the five centres were compared with gold standard MIC results that had been interpreted as susceptible, intermediate or resistant using the interpretive standards of the NCCLS; 12 agreement rates for individual antibiotics ranged from 88.9% to 100%. The agreements for organism groups were 98.0%, 95.8% and 94.4% for enterococci, staphylococci and Gram-negative bacilli, respectively (Table 3) with an overall mean of 95.1%. Very major, major and minor error rates for enterococci were 0.4%, 0.4% and 1.2%, respectively. Corresponding error rates for staphylococci were 0.9%, 2.8% and 0.5% with error rates for Gramnegative bacilli of 0.8%, 2.4% and 2.4%. Tables 4 6 indicate the degree of agreement between inferred resistance mechanisms and reference results and, since VITEK 2 AES data represent detection of phenotypes in at least four of five centres, also indicate inter-site reproducibility for this method. In terms of detecting resistant phenotypes, VITEK 2 AES achieved 100% agreement with reference data for enterococci, successfully detecting overproduction of PBP5, specific mechanisms of glycopeptide resistance (vana, B and C), high-level aminoglycoside resistance, tetracycline resistance and resistance to MLS B antibiotics. VITEK 2 detected vanc resistance in one strain of E. gallinarum that was not detected by any routine method; tetracycline resistance in all three strains of enterococci was detected by VITEK and VITEK 2 only. VITEK 2 AES achieved 83.3% agreement with reference data for staphylococci, successfully detecting penicillinase production in Staphylococcus haemolyticus, APH(2 )-AAC(6 ) aminoglycoside resistance in S. aureus, fluoroquinolone resistance in S. aureus due to various mechanisms and rifampicin resistance in S. aureus. It failed to detect methicillin resistance in one strain of S. aureus and one strain of Staphylococcus warneri out of a total of 13 methicillin-resistant staphylococci. Lincosamide nucleotidylation was detected in one strain of S. haemolyticus, and teicoplanin resistance in one strain of S. aureus (although these were suggested as one of two possible mechanisms). VITEK 2 AES achieved 68.2% agreement with reference data for Gram-negative bacilli. It detected specific aminoglycoside resistance in four strains of Enterobacteriaceae on two occasions and, as a possible mechanism, on two others; it detected penicillinase production in Gramnegative bacilli on each of 11 occasions. Of 11 ESBL-producing strains, VITEK 2 AES detected eight directly and a further two as possible resistance mechanisms. It failed to detect the 1194

5 VITEK 2 Advanced Expert System Table 2. Comparison of MICs determined by VITEK 2 with those determined by the agar-dilution reference method No. of VITEK 2 MICs within indicated log 2 Organism/antibiotic < >+2 Total Essential agreement (%) a Enterococci Benzyl penicillin Ampicillin Tetracycline Erythromycin Nitrofurantoin Vancomycin Teicoplanin Total Percentage Staphylococci Benzyl penicillin Oxacillin Gentamicin Tetracycline Erythromycin Clindamycin TMP-SMX Nitrofurantoin Ofloxacin Rifampicin Vancomycin Teicoplanin Total Percentage Gram-negative bacilli Ampicillin Co-amoxiclav Ticarcillin Piperacillin Cefalothin Cefoxitin Cefotaxime Ceftazidime Cefepime Imipenem Gentamicin Tobramycin Netilmicin Amikacin TMP-SMX Ofloxacin Ciprofloxacin Total Percentage TMP-SMX, trimethoprim sulfamethoxazole. a Percentage of essential agreement within +/ 1 dilution. 1195

6 Table 3. Interpretive category errors for enterococci, staphylococci and Gram-negative bacilli No. tests Concordance (%) Minor error (%) Major error (%) Very major error (%) centre centre centre centre centre Organism/antibiotic VITEK VITEK VITEK VITEK VITEK 2 Enterococci Benzyl penicillin Ampicillin Tetracycline Erythromycin Nitrofurantoin Vancomycin Teicoplanin High-level gentamicin Mean (%) Total (n) Staphylococci Benzyl penicillin Oxacillin Gentamicin Tetracycline Erythromycin Clindamycin Nitrofurantoin Ofloxacin Vancomycin Teicoplanin Mean (%) Total (n) J. Barry et al. Gram-negative bacilli Ampicillin Co-amoxiclav Piperacillin Cefalothin Cefoxitin

7 VITEK 2 Advanced Expert System Cefotaxime Ceftazidime Imipenem Gentamicin Tobramycin Amikacin TMP-SMX Ofloxacin Ciprofloxacin Mean (%) Total (n) Methods: centres 1 and 2, agar-incorporation breakpoints; centres 3 and 4, BSAC disc diffusion; centre 5, VITEK. VITEK 2 data comprise combined data from all five centres. TMP-SMX, trimethoprim sulfamethoxazole. ESBL-producing strain of P. aeruginosa, as did every routine method. Of six cephalosporinase-producing Enterobacteriaceae, VITEK 2 AES detected two directly and a further two as possible mechanisms. It detected the combination of cephalosporinase and porin alteration, missed by every routine method, but did not detect the plasmid-mediated cephalosporinase that was also missed by every routine method. It failed to detect an E. coli impermeability mutant and detected carbapenem resistance in P. aeruginosa due to impermeability (either directly or as an alternative mechanism) but did not detect carbapenem resistance due to carbapenemases in Enterobacteriaceae. It detected five of six quinolone-resistant strains of E. coli, only failing to detect resistance due to a gyra mutation in one strain; mutations due to nalb and nald, and a gyra mutation in a second strain were detected by VITEK 2 AES but missed by every routine method that detected gyra + pare mutations only. The only test centre to use an MIC-based technique routinely was centre 5, which used VITEK. However, direct comparisons of MICs determined by VITEK and VITEK 2 were not possible because dilution ranges of antibiotics in the two test cards differed. NCCLS interpretive standards 12 were used to categorize gold-standard MIC results for centre 5 (VITEK); with other centres, comparator results were categorized using BSAC criteria. 14,15 Agreement rates for individual organism antibiotic combinations determined using routine methodology ranged from 66.7% to 100% indicating localized problems. However, overall agreement rates for test centres were generally comparable to those with VITEK 2: ranges were % (mean 95.3%), % (mean 97.1%) and % (mean 93.5%) for enterococci, staphylococci and Gram-negative bacilli, respectively, with an overall agreement of 95.0%. Very major, major and minor error rates for enterococci were %, % and %. Corresponding error ranges for staphylococci were %, % and % with error rates for Gram-negative bacilli of %, % and %. VITEK and VITEK 2 interpret a number of antibiotic results as intermediate and, thus, the number of minor errors with these methods will always be greater than methods using BSAC criteria that, for the most part, do not recognize an intermediate category. The degree of agreement between resistance mechanisms inferred using routine methodology and reference results was 73.7%, 52.6% and 52.6% for enterococci (centres 1, 3 and 4, respectively) compared to 100% with VITEK 2. The degree of agreement for staphylococci was 76.7%, 56.7%, 66.7% and 60.0% (centres 1, 3, 4 and 5, respectively) compared with 83.3% with VITEK 2. In contrast to VITEK 2 AES, all routine methods detected methicillin resistance in all S. aureus strains and in the single strain of S. warneri and centre 4 detected lincosamide nucleotidylation in the strain of S. haemolyticus. No routine method detected teicoplanin resistance in S. 1197

8 Table 4. Agreement between reference genotype data and individual laboratory or VITEK 2 AES phenotypic interpretation for enterococci Agreement Partial agreement Disagreement No response No. of tests centre centre centre centre VITEK VITEK VITEK VITEK E. faecium, overproduction of PBP High-level aminoglycoside resistance E. faecalis resistance to gentamicin E. faecium resistance to streptomycin and kanamycin 1 1 nt E. gallinarum resistance to kanamycin 1 1 nt nt Specific glycopeptide resistance E. faecium vana 1 1 nt E. faecalis vanb 2 2 nt E. faecium vanb 2 2 nt E. gallinarum vanc 2 1 nt Tetracycline resistance E. faecalis 1 nt 1 nt E. faecium 1 nt 1 nt E. gallinarum 1 nt 1 nt Resistance to MLS antibiotics E. faecalis E. faecium J. Barry et al. Total (all strains, all mechanisms) % Rows shaded in grey summarize the unshaded blocks below them, categorizing the organisms into larger groups. VITEK data represent detection of phenotypes in at least four of five analyses. Centres 2 and 5 carried out phenotypic determination using VITEK 2 only; nt, appropriate antibiotics not tested.

9 Table 5. Agreement between reference genotype data and individual laboratory or VITEK 2 AES phenotypic interpretation for staphylococci Agreement Partial agreement Disagreement No response No. of tests centre centre centre centre VITEK VITEK VITEK VITEK S. haemolyticus, penicillinase Methicillin resistance (meca) S. aureus S. epidermidis S. warneri Aminoglycoside resistance S. aureus, APH(2 )-AAC(6 ) Resistance to MLS antibiotics S. haemolyticus, efflux S. haemolyticus, lincosamide nucleotidylation S. aureus resistance to fluoroquinolones gyra, parc nt parc nt gyra 1 1 nt efflux 1 nt parc + efflux 1 1 nt gyra, parc + efflux 1 1 nt S. aureus resistance to rifampicin Low-level High-level Teicoplanin resistance S. aureus nt VITEK 2 Advanced Expert System Total (all strains, all mechanisms) % Rows shaded in grey summarize the unshaded blocks below them, categorizing the organisms into larger groups. VITEK data represent detection of phenotypes in at least four of five analyses. Centre 2 carried out phenotypic determination using VITEK 2 only; nt, appropriate antibiotics not tested.

10 Table 6. Agreement between reference genotype data and individual laboratory or VITEK 2 AES phenotypic interpretation for Gram-negative bacilli Agreement Partial agreement Disagreement No response No. of tests centre centre centre centre VITEK VITEK VITEK VITEK Specific aminoglycoside resistance E. coli AAC(3)-I E. coli AAC(6 ) E. coli ANT(2 ) P. stuartii natural AAC(2 ) Gram-negative bacilli, penicillinase Acquired penicillinase K. oxytoca high-level K E. coli inhibitor-resistant penicillinase Gram-negative bacilli, ESBL Enterobacteriaceae ESBL (TEM or SHV) P. aeruginosa (TEM, SHV or PER) Enterobacteriaceae cephalosporinases E. coli cephalosporinase High-level cephalosporinase High-level cephalosporinase + porin alteration Plasmid-mediated cephalosporinase P. aeruginosa high-level resistance E. coli impermeability mutant Gram-negative bacilli, carbapenem resistance P. aeruginosa, impermeability Enterobacteriaceae, carbapenemase (IMP-1) E. coli, quinolone resistance nalb or nald 2 nt nt gyra gyra and pare J. Barry et al. Total (all strains, all mechanisms) % Rows shaded in grey summarize the unshaded blocks below them, categorizing the organisms into larger groups. VITEK 2 data represent detection of phenotypes in at least four of five analyses. Centre 2 performed phenotypic determination using VITEK 2 only; nt, appropriate antibiotics not tested.

11 VITEK 2 Advanced Expert System aureus. Agreement for Gram-negative bacilli was 68.2%, 34.1%, 52.3% and 22.7% (centres 1, 3, 4 and 5, respectively) with centre 1 performing as well as VITEK 2. Centres 1 and 4 detected one strain with high-level cephalosporinase; centre 1 further detected the E. coli impermeability mutant and carbapenemase production in Enterobacteriaceae resistance mechanisms that were not detected by VITEK 2 AES. Discussion VITEK 2 and AES have been evaluated in several countries We have carried out an evaluation in five UK laboratories and, in addition to comparing VITEK 2 MIC data with gold standard agar-dilution MIC data and assessing the ability of VITEK 2 AES to recognize resistance phenotypes, we have extended our study to compare VITEK 2 with routine antimicrobial susceptibility testing methods. A total of 5063 MICs were determined with VITEK 2 in the five test centres. In comparison with the agar-dilution reference method, VITEK 2 gave an overall essential MIC agreement of 96.5%, 96.7% and 95.5% for enterococci, staphylococci and Gram-negative bacilli, respectively. These data conform with those of other published studies 3,9 and meet FDA criteria for any new MIC test method. 19 Inter-laboratory variations were minimal. In general, for Enterobacteriaceae, MICs determined by VITEK 2 were higher than those determined using an agar-dilution method. Cantón et al. 8 reported that VITEK 2 read lower than the reference method (a microdilution method) but Leclercq et al. 9 also found VITEK 2 to read higher than an agar-dilution method. Several β-lactam agents (including piperacillin, cefalothin, ceftazidime and cefotaxime) are inoculum dependent in that the heavier the inoculum used, the higher the determined MIC becomes. 20 Thus, the use of a heavy inoculum in a broth test should better favour expression of resistance in comparison with an agardilution method and will account for relatively lower essential agreement values with these antibiotics in comparison with imipenem, for example. Leclercq et al. 9 noticed discrepancies with piperacillin, cefotaxime and ceftazidime with ESBLproducing strains and with cefalothin with penicillinaseproducing strains and these corresponded to inoculum effects. For enterococci, the lowest degree of agreement was with nitrofurantoin and this was also found by Leclercq et al. 9 Data were analysed by comparison with gold standard agar-dilution MIC data that had been categorized according to the relevant interpretive standards. Results for individual organism antibiotic combinations ranged from 88.9% (gentamicin versus Gram-negative bacilli) to 100%. With VITEK 2, the overall agreements for organism groups were 98.0%, 95.8% and 94.4% for enterococci, staphylococci and Gramnegative bacilli, respectively. The overall mean was 95.1%. Corresponding agreements for organism groups examined by routine methods were % (mean 95.3%), % (mean 97.1%) and % (mean 93.5%) with an overall mean of 95.0%. With routine methods, there was also a wider range of agreements for individual organism antibiotic combinations ( %) reflecting localized problems. For example, centre 3 had problems with gentamicin and staphylococci subsequently found to be due to a batch of defective antibiotic discs; centre 4 had problems with enterococci relating to vancomycin, and centre 1 had problems with Enterobacteriaceae and co-amoxiclav MICs falling close to the breakpoint. Centre 5 also had problems with enterococci and vancomycin, and with Gram-negative bacilli and both cefotaxime and tobramycin: minor errors were generated as the VITEK uses an intermediate category. Centre 4 used an Oxoid Aura image and this zone reader can interpret zones within 1 mm of the breakpoint as borderline susceptible or borderline resistant results interpreted using these data may thus improve the performance of the method. Some laboratories do not attempt to determine resistance mechanisms routinely from raw antibiotic susceptibility data. Other laboratories fail to detect mechanisms, often because insufficient or inappropriate antibiotics are tested as part of the standard antibiotic panel and this is reflected in the data generated during this study. Overall agreement of phenotypic interpretation with reference genotypic data for enterococci, staphylococci and Gram-negative bacilli using VITEK 2 was 100%, 83.3% and 68.2%, respectively. Corresponding values using routine methodologies were %, % and %. Essential agreement between categories of susceptibility determined by VITEK 2 and the routine methods was based on raw unexpertized data. However, most Expert systems are capable of editing antibiograms based on inferred mechanisms and, furthermore, are capable of artificial learning. The strains used in this study are not representative of strains encountered, as a matter of course, in routine microbiology laboratories as they were selected on the basis of their challenging resistance mechanisms. Nonetheless, VITEK 2 performed susceptibility tests accurately and the AES detected and interpreted resistance mechanisms appropriately. In this study, routine microbiology methods were shown to perform susceptibility tests with comparable accuracy to VITEK 2. Further, a conventional computerized expert system 5 or a knowledgeable operator, whilst not completely as effective as the AES, also carried out interpretative reading adequately. Routine methods are also cost-effective, flexible and, for the most part, standardized. However, VITEK 2 offers a standardized method ideally suited to laboratories lacking familiarity with, increasingly, myriad resistance mechanisms 4,7 and/or those not testing an appropriate number, or range, of antibiotics to detect resistant phenotypes using interpretative reading. 6 The main benefit of the adoption of VITEK 2 is likely to be the provision of rapid results that could bear both clinical and financial rewards

12 J. Barry et al. Acknowledgements We thank biomérieux for the loan of the VITEK 2 system and for the generous donation of reagents; biomérieux Application Specialists for their assistance; Gill Webb for organizing the study and Patrick Marchais for analysis of the data. We also thank other members of our own departments for their valuable technical assistance. References 1. Funke, G., Monnet, D., debernardis, C., von Graevenitz, A. & Freney, J. (1998). Evaluation of the VITEK 2 system for rapid identification of medically relevant gram negative rods. Journal of Clinical Microbiology 36, Jossart, M. F. & Courcol, R. J. (1999). Evaluation of an automated system for identification of Enterobacteriaceae and nonfermenting bacilli. European Journal of Clinical Microbiology and Infectious Diseases 18, Pérez-Vazquez, M., Oliver A., Sánchez del Saz, B., Loza, E., Baquero, F. & Cantón, R. (2001). Performance of the VITEK 2 system for identification and susceptibility testing of routine Enterobacteriaceae isolates. International Journal of Antimicrobial Agents 17, Livermore, D. M. (1995). β-lactamases in laboratory and clinical resistance. Clinical Microbiology Reviews 8, Courvalin, P. (1996). Interpretive reading of antimicrobial susceptibility tests. Clinical Microbiology and Infection 2, Suppl. 1, S Livermore, D. M., Winstanley, T. G. & Shannon, K. P. (2001). Interpretative reading: recognizing the unusual and inferring resistance mechanisms from resistance phenotypes. Journal of Antimicrobial Chemotherapy 47, Suppl. 1, Sanders, C. C., Peyret, M., Moland, E. S., Shubert, C., Thomson, K. S., Boeufgras, J.-M. et al. (2000). Ability of the VITEK 2 Advanced Expert System to identify β-lactam phenotypes in isolates of Enterobacteriaceae and Pseudomonas aeruginosa. Journal of Clinical Microbiology 38, Cantón, R., Pérez-Vázquez, M., Oliver, A., Coque, T. M., Loza, E., Ponz, F. et al. (2001). Validation of the VITEK2 and the Advanced Expert System with a collection of Enterobacteriaceae harbouring extended spectrum or inhibitor resistant β-lactamases. Diagnostic Microbiology and Infectious Disease 41, Leclercq, R., Nicolas-Chanoine, M.-H., Nordmann, P., Phillipon, A., Marchais, P., Buu-Ho, A. et al. (2001). Multicentre evaluation of an automated system using selected bacteria that harbour challenging and clinically relevant mechanisms of resistance to antibiotics. European Journal of Clinical Microbiology and Infectious Diseases 20, Sanders, C. C., Peyret, M., Moland, E. S., Cavalieri, S. J., Shubert, C., Thomson, K. S. et al. (2001). Potential impact of the VITEK 2 system and the Advanced Expert System on the clinical laboratory of a university-based hospital. Journal of Clinical Microbiology 39, Livermore, D. M., Struelens, M., Amorim, J., Baquero, F., Bille, J., Canton, R. et al. (2002). Multicentre evaluation of the VITEK 2 Advanced Expert System for interpretive reading of antimicrobial resistance tests. Journal of Antimicrobial Chemotherapy 49, National Committee for Clinical Laboratory Standards. (2000). Performance Standards for Antimicrobial Disk Susceptibility Tests Seventh Edition: Approved Standard M1000-S11. NCCLS, Villanova, PA, USA. 13. Faiers, M., George, R., Jolly, J. & Wheat, P. (1991). Multipoint Methods in the Clinical Laboratory, A Handbook. Public Health Laboratory Service, London, UK. 14. Andrews, J. M. for the BSAC Working Party on Susceptibility Testing. (2001). BSAC standardized susceptibility test method. Journal of Antimicrobial Chemotherapy 48, Suppl. 1, Andrews, J. (Ed.) (2002). BSAC Disc Diffusion Method for Antimicrobial Susceptibility Testing. August 2002, revision date [On-line.]. viewitem&itemid=102 (18 February 2003, date last accessed). 16. Andrews, J. M., Boswell, F. J. & Wise, R. (2000). Evaluation of the Oxoid Aura image system for measuring zones of inhibition with the disc diffusion technique. Journal of Antimicrobial Chemotherapy 46, Jarlier, V., Nicolas, M. H., Fournier, G. & Phillipon, A. (1988). Extended broad-spectrum β-lactamases conferring transferable resistance to newer β-lactam agents in Enterobacteriaceae: hospital prevalence and susceptibility patterns. Reviews of Infectious Diseases 10, Carter, M. W., Oakton, K. J., Warner, M. & Livermore, D. M. (2000). Detection of extended-spectrum β-lactamases in klebsiellae with the Oxoid combination disk method. Journal of Clinical Microbiology 38, Food and Drug Administration. (1991). Federal Guidelines. Review Criteria for Assessment of Antimicrobial Susceptibility Testing Device. Food and Drug Administration, Rockville, MD, USA. 20. Johnson, C. C., Livornese, L., Gold, M. J., Pitsakis, P. G., Taylor, S. & Levison, M. E. (1995). Activity of cefepime against ceftazidime-resistant gram-negative bacilli using low and high inocula. Journal of Antimicrobial Chemotherapy 35, Doern, G., Vautour, R., Gaudet, M. & Levy, B. (1994). Clinical impact of rapid in vitro susceptibility testing and bacterial identification. Journal of Clinical Microbiology 32,