Performance of Vitek 2 in Antimicrobial Susceptibility Testing of Pseudomonas aeruginosa Isolates with Different Mechanisms of β-Lactam Resistance

Abstract

A total of 78 isolates of Pseudomonas aeruginosa grouped according to the phenotype for ceftazidime and imipenem susceptibility/resistance were used to assess the accuracy of the Vitek 2 system in antimicrobial susceptibility testing. Comparisons were made with a MIC gradient test for piperacillin-tazobactam, ceftazidime, aztreonam, imipenem, meropenem, gentamicin, and ciprofloxacin. For the total of 546 isolate-antimicrobial combinations tested, the category agreement was 83.6%, with 2.0, 1.6, and 12.8% very major, major, and minor errors, respectively. Vitek 2 accuracy was influenced differently by the mechanism responsible for resistance, and interpretation of the results in relation to phenotype could improve the performance of the system.

Broth microdilution and disk diffusion are the reference methods for antimicrobial susceptibility testing (AST) of Pseudomonas aeruginosa to broad-spectrum β-lactam antibiotics, whereas there has been longstanding concern about the accuracy of commercial automated systems (6, 9, 10, 14). Since the automated systems are potentially capable of improving standardization, requiring less labor, and validating MIC results by providing expert systems, the poor performance they display in P. aeruginosa AST constitutes a major drawback for clinical microbiology laboratories. There is no obvious explanation for the discrepancy between the results obtained with automated systems and reference methods, but the importance of factors such as the phenotypic characteristics of the organisms and the underlying resistance mechanisms is well known (14). Several mechanisms may lead to resistance to broad-spectrum β-lactams: hyperproduction of broad-spectrum β-lactamases (including carbapenemases), decreased outer membrane permeability, and active efflux (5, 13, 17, 18).

In our study, we considered isolates with different phenotypes of susceptibility/resistance (each based on different biochemical mechanisms) to imipenem (Ipm) and ceftazidime (Caz) and evaluated the accuracy of a commonly used automated system, the Vitek 2 system (bioMérieux, Rome, Italy), for isolates belonging to each phenotype. Our intention was to produce information for reassessment of the β-lactam interpretative algorithms of this system for tests with P. aeruginosa.

Seventy-eight isolates of P. aeruginosa selected from our collection on the basis of AST performed by an agar diffusion technique were used in this study. Twenty-three were Ipm resistant (Ipm^r) and Caz susceptible (Caz^s) (group 1), 23 were Caz^r Ipm^s (group 2), 15 were Caz^r Ipm^r (group 3), and 17 were Caz^s Ipm^s (group 4). All strains were isolated over the period 2000 to 2006 from different patients and clinical settings at the University Hospital of Verona, Italy, and were unique pulsed-field gel electrophoresis types, without patient duplicates to minimize any clonal influence. In 2006, the distribution of these phenotypes among routine clinical isolates in our institution was 6.7% (group 1), 7.4% (group 2), 13.6% (group 3), and 67% (group 4). The remaining 5% were not named.

Susceptibility to piperacillin-tazobactam (Tzp), Caz, aztreonam (Azt), Ipm, meropenem (Mer), gentamicin (Gen), and ciprofloxacin (Cip) was determined by the Vitek 2 system by using the AST-N022 card (bioMérieux, Rome, Italy) and a MIC gradient test validated for testing P. aeruginosa (2) (Etest; AB Biodisk, Solna, Sweden) according to the manufacturer's instructions and the latest CLSI documents (3). Category agreement (CA) was based on the criteria proposed for large samples of known resistant isolates (8): very major interpretative category errors (VME) amounted to ≤3% and the combination of major interpretative category errors (ME) and minor interpretative category errors (mE) amounted to ≤7%, so that the total CA should be >90% for the accuracy to be acceptable. Essential agreement (EA) was considered when the Vitek 2 MIC was within a range of three twofold dilutions (target value ± 1 dilution).

AmpC cephalosporinase activities were determined in all isolates by spectrophotometric analysis using nitrocefin as a substrate. The presence of carbapenemase activity was investigated in group 1 and 3 isolates by measuring the hydrolysis rates of 100 μM Ipm in cell sonic extracts as previously described (4) and by amplification of bla_VIM and bla_IMP genes encoding the VIM and IMP metallo-β-lactamases, respectively (12). Overproduction of the MexAB-OprM efflux pump and OprD expression were assessed in Imp^r isolates negative for metallo-β-lactamase genes (all belonging to group 1) by Western blottingof MexB with anti-MexB rabbit antibodies (a generous gift from K. Poole). In each experiment, PAO1 (showing a normal expression level for the MexAB-OprM system) was included as a standard and the MexB level of each strain was calculated as a relative value in comparison with that of PAO1. OprD expression was evaluated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis of the outer membrane proteins (18).

The resistance/susceptibility pattern of isolates to the antimicrobial tested as determined by the MIC gradient test is reported in Table 1.

TABLE 1.

Susceptibility patterns of 78 P. aeruginosa isolates determined by MIC gradient test and grouped according to phenotype

Group (no. of isolates)	Phenotypea	No. of isolates with susceptibility patternb:
		Tzp	Caz	Azt	Ipm	Mer	Gen	Cip
1 (23)	Cazs Ipmr	23 S	23 S	18 S, 3 I, 2 R	23 R	4 S, 5 I, 14 R	11 S, 12 R	12 S, 11 R
2 (23)c	Cazr Ipms	10 S, 5 I, 8 R	4 I, 19 R	5 S, 11 I, 7 R	23 S	23 S	6 S, 1 I, 16 R	9 S, 14 R
3 (15)	Cazr Ipmr	2 I, 13 R	15 R	5 S, 3 I, 7 R	15 R	15 R	15 R	15 R
4 (17)	Cazs Ipms	17 S	17 S	17 S	17 S	17 S	16 S, 1 R	14 S, 4 R

^aAccording to the disk diffusion test.

^bAccording to the MIC gradient test. S, susceptible; I, intermediate; R, resistant.

^cAll nonsusceptible isolates were considered as Caz^r.

Group 1 included Caz^s Ipm^r phenotype isolates. All isolates were negative for carbapenemase activity and did not overproduce AmpC cephalosporinase. OprD protein was absent in all isolates, and MexB production was increased in all isolates (from 1 to 5 times). These results suggested alterations in permeability/efflux systems as the mechanisms responsible for the Ipm^r phenotype observed (permeability phenotype, according to reference 16). From a total of 161 isolate-antimicrobial combinations tested in the Vitek 2, the overall CA was 71.4%, with 1.8, 2.4, and 24.2% VME, ME, and mE, respectively. An acceptable level of accuracy was evidenced for Tzp and Cip. An unacceptable level of accuracy was evidenced for Caz, Azt, Ipm, Mer, and Gen. For Caz and Gen, the rate of mE (the only type of error detected) reflected the number of isolates whose MICs were within ±1 log₂ dilution of the categorical breakpoints and the EA for these antimicrobial agents was higher than 90% since for most isolates it might be possible to spuriously categorize a 1-dilution MIC error as an mE. The tendency of the Vitek results was toward increased MICs for Azt (13% ME) but toward lower MICs for Mer (13% VME).

Group 2 isolates were all nonsusceptible to Caz (19 resistant and 4 intermediate) but susceptible to Ipm (Caz^r Ipm^s phenotype). Overproduction of cephalosporinase activity was evidenced in all isolates (high-level cephalosporinase phenotype, according to reference 16). No attempt was made to characterize the enzymes responsible for such activity (the chromosomally encoded AmpC cephalosporinases and/or the concomitant presence of plasmid-encoded secondary β-lactamase), since this was beyond the aim of our study. On a total of 161 antimicrobial-microorganism combinations, the overall CA was 82.6% with 2.5, 0.6, and 14.3% VME, ME, and mE, respectively. An acceptable level of accuracy was evidenced for Ipm, Mer, and Cip. An unacceptable level of accuracy was evidenced for Tzp, Caz, Azt, and Gen. For Azt and Gen, the rate of mE reflected the number of isolates whose MICs were within ±1 log₂ dilution of the categorical breakpoints and the EA for these antimicrobial agents was higher than 90%. The tendency of the Vitek 2 results was toward increased MICs for Azt (4.3% ME) but toward lower MICs for Tzp (17.3% VME).

Group 3 isolates were all resistant to both antimicrobial agents (Caz^r Ipm^r phenotype) and produced a VIM-1 metallo-β-lactamase. Efflux pump overexpression and lack of OprD porin were not investigated in these isolates since these mechanisms do not contribute to resistance to broad-spectrum β-lactams other than carbapenems. For a total of 105 antimicrobial-microorganism combinations, the overall category agreement was 88.6%, with 3.8, 3, and 5% VME, ME, and mE, respectively. An acceptable level of accuracy was evidenced for Caz, Ipm, Mer, Gen, and Cip. An unacceptable level of accuracy was evidenced for Tzp and Azt. The tendency of the Vitek 2 results was toward increased MICs for Azt (20% ME) but toward lower MICs for Tzp (26.6% VME).

All group 4 isolates showed a wild-type phenotype (Caz^s Ipm^s), and none displayed significant β-lactamase activity. For a total of 119 antimicrobial-microorganism combinations, the overall category agreement was 94.1%, with 0.8 and 5% ME and mE, respectively. An acceptable level of accuracy was evidenced for Tzp, Caz, Ipm, Mer, and Gen. An unacceptable level of accuracy was evidenced for Azt and Cip. For Cip, the rate of mE reflected the number of isolates whose MICs were within ±1 log₂ dilution of the categorical breakpoints and the EA for this antibiotic was 100%.

For the total of 546 isolate-antimicrobial combinations tested, the category agreement was 83.6% with 2.0, 1.6, and 12.8% VME, ME, and mE, respectively.

AST of all strains for which VME and ME were detected (seven in group 1, five in group 2, seven in group 3, and one in group 4, making 20 isolates in all and with 140 isolate-antimicrobial combinations) was repeated in triplicate with both methods to assess the reproducibility of the errors. Acceptable reproducibilities (MICs within 3 log₂ dilutions) were 97.7% for Vitek and 96.5% for the MIC gradient test. CA was confirmed in 75.2% of Vitek 2 repeat testing and in 89.0% of the MIC gradient repeat testing (not shown). Table 2 gives the numbers of VME and ME resolved or confirmed or changed to mE upon repeat testing. On considering the VME and ME resolved in repeat testing, the VME and ME rates of Vitek 2 fell to 1.5 and 0.2%, respectively.

TABLE 2.

Accuracy of Vitek 2 for P. aeruginosa AST

Group (no. of isolates)	Agent	EA (%)	CA (%)	No. (%) of errors			VME and ME outcome upon repeat testing
				VME	ME	mE
1 (23)	Tzp	91.3	100	0	0	0
	Caz	95.6	56.5	0	0	10 (56.5)
	Azt	60.9	47.8	0	3 (13)	9 (39.1)	All ME changed to mE
	Ipm	69.5	65.2	0	0	8 (34.7)
	Mer	78.2	65.2	3 (13)	0	5 (21.7)	2 VME changed to mE, 1 resolved
	Gen	95.6	78.2	0	0	5 (21.7)
	Cip	91.3	95.6	0	1 (4.3)	0	ME resolved
2 (23)	Tzp	60.8	60.8	4 (17.3)	0	5 (21.7)	All VME confirmed
	Caz	86.9	86.9	0	0	3 (13)
	Azt	91.3	52.2	0	1 (4.3)	10 (56.5)	ME resolved
	Ipm	95.6	100	0	0	0
	Mer	73.9	100	0	0	0
	Gen	100	78.2	0	0	5 (21.7)
	Cip	100	100	0	0	0
3 (15)	Tzp	73.3	60	4 (26.6)	0	2 (13.3)	All VME confirmed
	Caz	100	100	0	0	0
	Azt	80	73.3	0	3 (20)	1 (0.6)	All ME changed to mE
	Ipm	100	100	0	0	0
	Mer	93.3	93.3	0	0	1 (0.6)
	Gen	100	100	0	0	0
	Cip	100	100	0	0	0
4 (17)	Tzp	70.5	100	0	0	0
	Caz	94.1	100	0	0	0
	Azt	64.7	76.4	0	0	4 (23.5)
	Ipm	94.1	94.1	0	1 (5.8)	0	ME confirmed
	Mer	94.1	100	0	0	0
	Gen	94.1	100	0	0	0
	Cip	100	88.2	0	0	2 (11.6)
Total (78)	Tzp	74.3	80.7	8 (10.2)	0	7 (8.9)	All VME confirmed
	Caz	93.5	83.3	0	0	13 (16.6)
	Azt	74.3	60.2	0	7 (8.9)	24 (30.7)	6 ME changed to mE, 1 resolved
	Ipm	88.4	88.4	0	1 (1.3)	8 (10.2)	ME confirmed
	Mer	83.3	88.4	3 (3.8)	0	6 (7.6)	2 VME changed to mE, 1 resolved
	Gen	97.4	87.1	0	0	10 (12.8)
	Cip	97.4	96.1	0	1 (1.3)	2 (2.5)	ME resolved
Total combinations (546)		87.0	83.6	11 (2.0)	9 (1.6)	70 (12.8)

Several studies on the accuracy of Vitek 2 have been conducted, but results have proven difficult to compare because they were carried out either on the performance of only one or two β-lactams (7, 11), on very few isolates (16), or on isolates from cystic fibrosis patients (1). Our results showed good agreement with previous studies carried out with Vitek 2 and a substantial number of P. aeruginosa isolates (9, 10, 14). Systematic biases toward false susceptibility to Tzp and toward false resistance to Azt have also been evidenced by others with the Vitek 2 system (10, 14), whereas false susceptibility to Ipm and false resistance to Caz, also detected by others (10, 14), were not detected in our study.

Acceptable levels of accuracy for Caz and Cip were confirmed by Saegeman et al. (15). To the best of our knowledge, this study was the first aimed at evaluating Vitek 2 accuracy according to the particular resistance phenotype of the strains. Altered permeability (group 1) seemed to play an important role in decreasing the accuracy not only against carbapenems, but also against Caz, Azt, and Gen. Tzp and Cip were less affected by permeability defects. False susceptibility to Mer was detected only in this group (three isolates, one intermediate and two resistant to Imp). However, the simultaneous testing of both carbapenems should help the microbiologist and the clinician identify possible problems with carbapenem susceptibility. High cephalosporinase and metallo-β-lactamase levels were consistently associated with the significant bias toward false susceptibility for Tzp, as VME for this antibiotic were detected only in groups 2 and 3. This result underlines the need to insert a new rule in the expert system that excludes the Tzp result in the interpretation of the AST of isolates with these phenotypes.

The trend toward increased MICs for Azt reported previously (10) was confirmed in our study and generated both ME and mE in all groups. From the clinical point of view, this inaccuracy did not appear relevant for isolates of groups 2 and 3, since Azt is not considered a safe alternative for the therapy of infections due to Caz^r group 2 isolates, but such errors would limit the use of a usable agent in the case of infections due to group 1 and 4 isolates.

The trend of mE of Caz was consistently toward higher MICs (intermediate versus susceptible) in group 1 and lower MICs (intermediate versus resistant) in group 2. The trend of Ipm was toward lower MICs (intermediate versus resistant) in group 1. The trend of Gen was toward higher MICs (intermediate versus susceptible) in groups 1 and 2. mE were lower for group 3, indicating that the Vitek 2 system affords greater reliability for organisms with high levels of resistance.

In conclusion, in spite of the overall CA below the acceptable level (mainly generated by mE), our study indicates that Vitek 2 could be used with confidence for identifying resistance to several clinically important antimicrobial agents (Caz, Imp, Gen, and Cip). Once the interpretative algorithms of this system for tests with P. aeruginosa have been reassessed and the biases detected and corrected and once the various types of errors detected have been minimized or eliminated, alternative methods for routine AST of P. aeruginosa isolates based on validated manual methods (disk diffusion and MIC gradient) could be limited to isolates from serious infections, thus reducing the drawback of excluding automated systems for the AST of all P. aeruginosa isolates in clinical microbiology laboratories.