Abstract

MATERIALS AND METHODS

RESULTS

Bacterial ID.

Table ​Table11 shows the results of testing of 150 isolates of Enterobacteriaceae and 45 NFGNB. The Phoenix and MicroScan systems showed percentages of agreement at the genus and species levels of 98.7 and 97.7%, respectively, for the various isolates of Enterobacteriaceae. Two isolates of the Enterobacteriaceae family required the resolution of results by testing with the API 20E. One isolate was initially identified by the MicroScan system as Yokenella (Koserella) regensburgei and as Hafnia alvei by the Phoenix system. The MicroScan system identified the second isolate as Serratia fonticola or Enterobacter aerogenes, versus the Phoenix ID as Enterobacter aerogenes or Enterobacter cloacae. Following confirmatory testing, the Phoenix system was found to have correctly identified both isolates with discrepant results, which gave an overall accuracy of the Phoenix system of 100% for the ID of Enterobacteriaceae to the genus and species levels.

TABLE 1.

ID results for Enterobacteriaceae and nonfermenters

Organism	No. of isolates	% Of isolates for which Phoenix ID was:		API 20E ID
		Concordant with MicroScan ID	Discordant with MicroScan ID
Enterobacteriaceae
Klebsiella pneumoniae	45	100	0
E. coli	25	100	0
Enterobacter cloacae	25	100	0
Proteus mirabilis	18	100	0
Enterobacter aerogenes	11	100	0
Serratia marcescens	10	100	0
Klebsiella oxytoca	9	100	0
Proteus vulgaris	1	100	0
Morganella morganii	1	100	0
Salmonella sp.	1	100	0
Citrobacter freundii	1	100	0
Pantoea (Enterobacter) agglomerans	1	100	0
Yokenella (Koserella) regensburgei	1	0	100	Hafnia alveia
Serratia fonticola or Enterobacter aerogenes	1	0	100	Enterobacter aerogenesb
Total	150
Nonfermenters
Pseudomonas aeruginosa	26	96.2	3.8	Mucoid Pseudomonas aeruginosac
Acinetobacter baumannii	8	100	0
Stenotrophomonas maltophilia	8	100	0
Chromobacterium violaceum	1	100	0
Moraxella sp.	1	0	100	Kingella kingaed
Pseudomonas fluorescens or Pseudomonas putida	1	100	0
Total	45

Open in a separate window

^aPhoenix ID: Hafnia alvei.

^bPhoenix ID: Enterobacter aerogenes or Enterobacter cloacae.

^cPhoenix ID: Pseudomonas putida or Pseudomonas aeruginosa (Pseudomonas fluorescens or Pseudomonas putida upon repeat testing).

^dPhoenix ID: Kingella kingae.

For the 45 NFGNB isolates, two discrepant IDs by the Phoenix system were noted. One isolate, a mucoid strain of Pseudomonas aeruginosa, was misidentified as Pseudomonas putida or Pseudomonas aeruginosa and Pseudomonas fluorescens or Pseudomonas putida. Following confirmatory testing, the isolate was identified as Pseudomonas aeruginosa. The second discrepant isolate, following confirmatory testing, was found to have been correctly identified by the Phoenix system as Kingella kingae, versus an incorrect ID by the MicroScan system as Moraxella sp. The overall accuracy of the Phoenix system for the ID of NFGNB was 100% at the genus level and 97.7% at the species level.

Each system had difficulty in providing precise IDs of the eight challenge NFGNB isolates (Table ​(Table2).2). Both systems correctly identified the two strains of Ralstonia mannitolilytica to the genus level but misidentified both strains as Ralstonia pickettii. Of the five Burkholderia isolates, the MicroScan and Phoenix systems correctly identified one and three isolates, respectively, to the genus level; all five isolates were misidentified at the species level by both systems. The Phoenix system yielded a “no ID” message for the strains of Burkholderia multivorans and Burkholderia cenocepacia, whereas the MicroScan system generated a “no ID” result for two strains of Burkholderia cenocepacia. Pandoraea pnomenusa was incorrectly identified as Alcaligenes xylosoxidans and Moraxella sp. by the MicroScan and Phoenix systems, respectively. Burkholderia gladioli was misidentified by the MicroScan system as Acinetobacter baumannii, while the Phoenix system identified the isolate as Burkholderia sp. or Ralstonia sp.

TABLE 2.

Summary of ID results for challenge isolates

MicroScan ID (% probability)	Phoenix ID	Reference ID
Ralstonia pickettii	Ralstonia pickettii	Ralstonia mannitolilytica
Ralstonia pickettii (51) or Burkholderia cepacia (42)	Burkholderia cepacia or Ralstonia pickettii	Ralstonia mannitolilytica
Burkholderia cepacia	No ID	Burkholderia multivorans
Alcaligenes xylosoxidans	Moraxella sp.	Pandoraea pnomenusa
No ID	Burkholderia cepacia	Burkholderia cenocepacia
Vibrio sp. or Chryseobacterium sp.	Burkholderia cepacia	Burkholderia cenocepacia
No ID	No ID	Burkholderia cenocepacia
Acinetobacter baumannii	Burkholderia sp. or Ralstonia sp.	Burkholderia gladioli

Open in a separate window

Antimicrobial susceptibility.

Antimicrobial susceptibility results for the Enterobacteriaceae and NFGNB isolates tested against individual antibiotics are summarized in Table ​Table3.3. Category agreement (CA) was defined as matching classifications as susceptible, intermediate, or resistant by the two systems. The level of CA for the Enterobacteriaceae was 96.5%. The mE, ME, and VME rates were 2.4, 1.4, and 6.0%, respectively. The majority of VMEs were associated with β-lactam antibiotics. β-Lactam antibiotics also accounted for the majority of MEs. Based on the indications of the MicroScan ESBL screening software (version 2.0), nine clinical isolates (seven Klebsiella pneumoniae and two E. coli isolates) were further screened as possible ESBL producers (Table ​(Table4).4). Following confirmatory testing, six Klebsiella pneumoniae isolates were determined to be negative for ESBL production and one was found to be positive; both E. coli isolates were negative. The Phoenix system correctly detected the one ESBL-producing Klebsiella pneumoniae isolate and incorrectly detected one of the two E. coli isolates as an ESBL-producing organism (Table ​(Table4).4). Of the 10 previously characterized ESBL-producing challenge isolates (5 Klebsiella pneumoniae, 3 Klebsiella oxytoca, and 2 E. coli isolates), the Phoenix instrument correctly identified 6 isolates, with a “no alert” message recorded for 4 isolates; the MicroScan system correctly classified all 10 isolates as possible ESBL producers with the message “?ESBL,” indicating the need for confirmatory testing.

TABLE 3.

AST results for Enterobacteriaceae and gram-negative nonfermenters

Organism group and antibiotic	Total no. of isolates tested	% CA	No. of isolates categorized asa:			No. (%) of isolates associated with:
			S	I	R	mE	ME	VME
Enterobacteriaceae
Amikacin	132	100.0	132	0	0	0	0	0
Gentamicin	131	98.0	124	4	3	3 (2.3)	0	0
Tobramycin	135	96.0	124	7	3	4 (2.9)	0	0
Imipenem	119	100.0	119	0	0	0	0	0
Meropenem	131	100.0	131	0	0	0	0	0
Cephalothin	83	83.8	65	2	16	7 (8.4)	0	1 (6.2)
Ceftazidime	132	95.9	105	3	24	2 (1.5)	1 (0.9)	2 (8.3)
Cefotaxime	133	96.6	109	7	17	3 (2.3)	0	1 (5.8)
Cefepime	114	96.9	110	1	1	2 (1.7)	0	1 (33.3)
Aztreonam	132	94.6	114	2	2	3 (2.3)	3 (2.6)	2 (1.4)
Piperacillin	114	94.6	77	11	11	7 (6.1)	0	0
Amoxicillin-clavulanic acid	123	90.6	73	8	8	8 (6.5)	0	0
Trimethoprim-sulfamethoxazole	124	99.3	108	0	0	0	0	1 (6.2)
Ciprofloxacin	99	98.3	89	1	1	2 (2.0)	0	0
Gatifloxacin	132	98.0	121	1	1	2 (1.5)	1 (0.8)	0
Levofloxacin	132	98.6	121	0	0	2 (1.5)	0	0
Tetracycline	75	98.9	67	1	1	1 (1.3)	0	0
Total	2,041	96.5	1,789	48	48	46 (2.4)	5 (1.4)	8 (6.0)
Nonfermenters
Amikacin	35	91.4	31	4	0	3 (8.6)	0	0
Gentamicin	35	82.9	26	5	4	5 (14.2)	1 (2.9)	0
Tobramycin	35	94.3	32	1	2	2 (5.7)	0	0
Imipenem	35	100.0	32	0	4	0	0	0
Meropenem	35	100.0	32	0	3	0	0	0
Ceftazidime	35	97.1	32	1	2	1 (2.9)	0	0
Cefepime	26	84.6	19	6	1	4 (15.4)	0	0
Piperacillin	35	94.3	29	2	4	2 (5.7)	0	0
Ciprofloxacin	35	97.1	25	0	10	1 (2.9)	0	0
Levofloxacin	42	90.5	30	2	10	4 (9.5)	0	0
Total	348	93.2	288	21	40	22 (8.0)	1 (2.9)	0

Open in a separate window

^aS, susceptible; I, intermediate; R, resistant.

TABLE 4.

Detection of ESBL-producing challenge and clinical isolates

Organism	Resulta from:		Disk confirmation	ESBL	Other β-lactamase(s)
	MicroScan	Phoenix
Clinical strains
Klebsiella pneumoniae	?ESBL	No alert	ND
	?ESBL	No alert	ND
	?ESBL	No alert	Negative
	?ESBL	No alert	Negative
	?ESBL	No alert	Negative
	?ESBL	No alert	Negative
	?ESBL	1505	Positive
E. coli	?ESBL	1505	Negative
	?ESBL	No alert	Negative
Challenge strains
Klebsiella pneumoniae	?ESBL	1505		SHV-3-like ESBL	SHV-1-like β-lactamase
	?ESBL	1505		CTX-M-19	SHV-1-like and TEM-1-like β-lactamases
	?ESBL	No alert		SHV-4-like ESBL	PSE-like, FOX-like AmpC, and SHV-1-like β-lactamases
	?ESBL	No alert		SHV-5 and SHV-3	ACT-1 AmpC, SHV-1, and TEM-1
	?ESBL	No alert		SHV-3-like ESBL	SHV-1-like β-lactamase
Klebsiella oxytoca	?ESBL	No alert		SHV-3-like ESBL	DHA-like AmpC, K1, and TEM-1-like β-lactamases
	?ESBL	1505		SHV-1-like ESBL	K1 and TEM-1-like β-lactamases
	?ESBL	1505		TEM-3-like ESBL	K1
E. coli	?ESBL	1505		TEM-12	TEM-1
	?ESBL	1505		TEM-10
Serratia marcescens	?ESBL1	No alert		SHV-4-like β-lactamase	AmpC (chromosomal)
Enterobacter cloacae	?ESBL1	No alert		SHV-5-like β-lactamase	AmpC (chromosomal) and TEM-1
Citrobacter koseri	?ESBL1	No alert		SHV-7	TEM-1-like and OXA-like β-lactamases
Proteus mirabilis	?ESBL1	No alert		CTX-M-2	TEM-1-like β-lactamase

Open in a separate window

^a?ESBL, positive ESBL screen as defined by software using CLSI guidelines (no confirmation available on panel); ?ESBL¹, positive ESBL screen as defined by user (no confirmation available on panel); 1505, Phoenix system alert for a confirmed ESBL producer; ND, not done.

Overall complete agreement between the two systems following AST of NFGNB was 93.2%; the mE, ME, and VME rates were 8.0, 2.9, and 0%, respectively. Gentamicin and trimethoprim/sulfamethoxazole accounted for the two MEs.

DISCUSSION

Overall, the Phoenix instrument compared favorably to the MicroScan system for the ID and susceptibility testing of gram-negative bacilli. We purposely restricted the number of E. coli and Pseudomonas aeruginosa isolates to ensure a balanced representation of clinical isolates, especially those with low frequencies of occurrence in our laboratory. Although the range and number of isolates tested were not as broad as those described in comparable published reports, they do reflect a realistic representation of the organisms encountered by a service-oriented clinical microbiology laboratory. The overall agreement for the ID of common members of the family Enterobacteriaceae was similar to that reported by others, although in the previously reported studies, the performance of the Phoenix instrument was compared directly to conventional ID systems (1, 2, 12). In contrast, O'Hara tested a set of challenge isolates consisting of biochemically typical and atypical stock culture representatives of the Enterobacteriaceae and reported that the Phoenix system correctly identified only 89.9% to the genus and species levels (13). The majority of errors reported in her study were associated with the genus Salmonella, but no such errors were encountered in the present study. In the study by Carroll et al. (2), an Enterobacter (Pantoea) species, specifically Pantoea agglomerans, was misidentified as Shigella flexneri, one isolate of E. coli was misidentified as Salmonella sp., and all Shigella isolates were misidentified as E. coli. In our study, only two isolates required the resolution of results with the API 20E, and in these cases, the MicroScan system misidentified Hafnia alvei as Yokenella (Koserella) regensburgei and Enterobacter aerogenes as Serratia fonticola or Enterobacter aerogenes. Following the resolution of the two discrepancies, the Phoenix instrument was found to have achieved an overall accuracy of ID of Enterobacteriaceae at the genus and species levels of 100%. However, it is important for microbiologists to be aware of the shortcomings of both automated and nonautomated ID systems, especially for isolates identified as Shigella or Salmonella spp., which require confirmatory ID by additional testing (agglutination or the detection of characteristic phenotypic features following growth on MacConkey agar), in addition to knowledge of the source of the specimen.

Following the resolution of two discrepancies, the Phoenix system was found to have accurately identified all of the NFGNB clinical isolates, including the Kingella kingae isolate, which was misidentified as Moraxella sp. by the MicroScan system, and the mucoid strain of Pseudomonas aeruginosa, even though Pseudomonas putida was part of the selection. The results for the ID of this group of organisms are similar to those reported by Endimiani et al. (7), Menozzi et al. (10), and Stefaniuk et al. (16). In contrast, O'Hara reported that of 57 isolates of NFGNB, 48 (84.2%) were correctly identified by the Phoenix system to the genus and species levels (12).

Both systems encountered considerable difficulty in identifying unusual isolates that are occasionally recovered, albeit in low frequencies, from cystic fibrosis patients, including Burkholderia species other than Burkholderia cepacia and Ralstonia and Pandoraea spp. Given that these organisms are rare and that Pseudomonas aeruginosa is the most common bacterium isolated from cystic fibrosis patients, these results are not surprising since these organisms are not included within the databases of the respective instruments. In fairness to both systems, Burkholderia cenocepacia is part of the Burkholderia cepacia complex and the organisms of this complex are difficult to distinguish biochemically. Other investigators have reported similar observations and concerns in utilizing automated systems for the ID of such organisms and have emphasized the importance of confirmatory testing of these traditionally difficult-to-identify organisms (1, 7, 12, 19). Our data also illustrate the potential problems that may be encountered when using these systems alone for the ID of unusual NFGNB and the need to perform confirmatory testing using more reliable phenotypic or genotypic methods.

In our study comparing AST results from the MicroScan and Phoenix systems, the CAs for both groups of organisms were excellent. A significant weakness of this study was the lack of sufficient numbers of resistant isolates to accurately determine the true rates of MEs and VMEs. In a service-oriented clinical microbiology laboratory in which clinical isolates are recovered from clinical specimens, like the one in which this study was conducted, some types of bacterial resistance occur at very low frequencies, which compromises the ability to generate data that are essential for calculating accurate error rates. This problem is reflected in the VME rates that exceed the accepted standard of ≤3% (6). Of particular note was the observation that the majority of VMEs involved β-lactam antibiotics. Menozzi et al. reported similar observations and noted, as was observed in our study, nearly equal numbers of errors for the various organisms tested (10). Similar findings were also reported by Carroll et al. for the Enterobacteriaceae (2).

Currently, five studies have extensively evaluated the ability of the Phoenix system to detect ESBLs (2, 10, 16, 17, 20). In the most recent study, Wiegand et al. (20) performed a side-by-side comparison of the MicroScan, Vitek 2, and Phoenix systems by testing 150 putative ESBL producers that were blindly distributed to three participating laboratories following the biochemical and molecular characterization of β-lactamases, which served as the reference method. A total of 85 isolates were identified as ESBL producers by the reference method. The Phoenix system demonstrated the highest sensitivity (99%) for the detection of ESBLs, followed by the Vitek 2 (86%) and the MicroScan (84%) systems. Carroll et al. demonstrated that the Phoenix expert rules system correctly detected the six ESBL producers, three E. coli and three Klebsiella spp. isolates, encountered in their study (2). Sturenburg et al. (17) also reported a high degree of sensitivity (100%) for the Phoenix system following the testing of 34 ESBL-producing challenge strains. Similar observations were reported by Sanguinetti et al. (14), who determined the Phoenix system to have 100% sensitivity and 98.9% specificity for the detection of ESBLs in 11 species of Enterobacteriaceae that had been characterized previously by phenotypic and genotypic methods. In a study of 74 ESBL-producing clinical isolates of E. coli and Klebsiella pneumoniae and 17 genotypically characterized control strains, Leverstein-van Hall et al. (9) compared the Phoenix system to the Vitek 1 and Vitek 2 systems, with the accuracies of the three systems for detecting ESBLs reported to be 89, 78, and 83%, respectively. In our study, both systems performed very well in screening for and/or detecting ESBL production in clinical isolates of Klebsiella pneumoniae and E. coli. However, the MicroScan system demonstrated higher levels of sensitivity and specificity in screening the ESBL-producing challenge isolates of E. coli, Klebsiella pneumoniae, and Klebsiella oxytoca that contained other β-lactamases. The higher-level performance of the Phoenix system reported by Wiegand et al. (20) and Sanguinetti et al. (14) was most likely due to the availability in Europe of expanded expert rules software (Phoenix software version 4.05W and MicroScan software version 1.12) that included organisms other than E. coli and Klebsiella pneumoniae, which is not available in the United States and is awaiting FDA clearance and CLSI recommendations for the testing of other members of the Enterobacteriaceae family, with the exception of the recent addition of guidelines for Proteus mirabilis.

Overall, the Phoenix system compared favorably with the MicroScan WalkAway system and is an acceptable alternative for the ID and susceptibility testing of commonly encountered species of Enterobacteriaceae and non-glucose-fermenting gram-negative bacilli. However, caution should be used in relying on either system for use in the ID and susceptibility testing of less commonly encountered clinical isolates, such as those that are occasionally recovered from patients with cystic fibrosis.

Acknowledgments

Footnotes

REFERENCES