INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Non-glucose-fermenting gram-negative bacilli are widely distributed in nature and are found in soil, water, and the general environment. They are opportunistic pathogens and are recovered from a variety of clinical specimens. Up to 20% of all gram-negative bacilli isolated in our hospital are non-glucose-fermenting bacilli. These organisms may cause community- and hospital-acquired infections either by colonization or accidental introduction into the body, or both (5). Some tend to be drug resistant (4); hence, a rapid, accurate, and cost-effective means of identification is required by laboratories (1, 3, 6, 7, 9, 10).

The identification of non-glucose-fermenting gram-negative bacilli by conventional biochemical methods is technical and labor-intensive (6, 7, 8, 9, 10, 11). Automated systems, such as the autoSCAN-W/A (W/A) (Dade Behring MicroScan Inc., West Sacramento, Calif.) and the Vitek AutoMicrobic System (AMS) (bioMérieux Vitek Systems Inc., Hazelwood, Mo.), provide fully automated microbiological diagnostic capabilities. These two systems offer greater standardization and convenience than the conventional identification systems, and they also shorten hands-on time (6, 7, 10, 11). In this study, 301 nonglucose-fermenting gram-negative bacilli were studied by two automatic systems, the W/A, with conventional-panel Neg Combo type 12, and the Vitek GNI+ AMS, and the results were compared with those of conventional biochemical methods for accuracy.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Isolates. A total of 301 non-glucose-fermenting gram-negative bacilli were collected from clinical specimens, representing 20 genera and 25 species. All isolates were tested by both the W/A conventional-panel Neg Combo type 12 and Vitek GNI+ identification systems. All isolates were stocked at 70°C and subcultured three times on Trypticase soy agar with 5% sheep blood (TSA II; Becton-Dickinson Microbiology System) at 35°C overnight before being tested. Definitive identification of all isolates was performed by conventional biochemical tests following the methodology published by the Centers for Disease Control and Prevention (12).

Identification system. The W/A automatic system with conventional-panel Neg Combo type 12 can identify non-glucose-fermenting gram-negative bacilli in 15 to 48 h. The oxidase test had been performed prior to inoculation on the autoSCAN-W/A system. Its concentration was standardized by Prompt Inoculation System-D, which contains the Prompt Inoculation wand with a collar and the Prompt Inoculation bottle with 30 ml of aqueous PLURONIC surfactants. Three to five isolated colonies were touched with the wand tip. Then the collar was pulled off and the wand was placed in the inoculation bottle. The bottle was shaken vigorously 8 to 10 times to release the bacteria from the wand tip. The final concentration of the bacterial suspension was 6.9 × 10⁵ CFU/ml. The standardized bacterial suspension was poured into a seed tray and passed through to the panel well. The labeled inoculated panel was inserted into the W/A incubation tower after oil was added to the designated wells and was then read automatically by its computer with version 22.01 software.

Definitions. In our research, we defined the term "correct at initial testing" to mean that the isolate was correctly identified to the species level at initial testing with a high probability of 85% by the W/A or 90% by the Vitek GNI+. The term "correct after additional testing" meant that additional testing recommended by the manufacturer's protocol was needed for correct identification when results were below the high probability. The term "error" meant that the isolate was misidentified by the system. All of the tests were repeated if the results were not correct at the initial testing.

Statistical analysis. The results were evaluated by chi-square analysis, using a t test to arrive at a P value.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

We tested 301 non-glucose-fermenting gram-negative bacilli representing 20 genera and 25 species using the W/A and Vitek GNI+ automatic microbiology systems. Tables 1 and 2 show the results of the W/A and Vitek GNI+ testing. A total of nine species were correct after additional testing in both systems: Acinetobacter baumannii, Alcaligenes xylosoxidans subsp. xylosoxidans, Bordetella bronchiseptica, Chryseobacterium meningosepticum, Ochrobactrum anthropi, Pseudomonas aeruginosa, Ralstonia pickettii, Shewanella putrefaciens, and Stenotrophomonas maltophilia. Four species were misidentified in both systems: Chryseobacterium indologenes, Pseudomonas fluorescens, Pseudomonas stutzeri, and Sphingomonas paucimobilis.

In the W/A system, 215 isolates (71.4%) were correctly identified to the species level at the initial testing, with a high probability of 85%. Another 74 isolates (24.6%) were identified correctly to the species level after additional testing, with a low probability of <85%, and the overall accuracy improved to 96.0%. In this study, we used an updated W/A conventional-panel database with revised software. Our results were better than those of Tenover et al. (11).

The Vitek GNI+ correctly identified 216 isolates (71.8%) to the species level at initial testing, with a high probability of 90%. Another 62 isolates (20.6%) were correct to the species level after additional testing, with a low probability of <90%. For these 62 isolates, the results were 55 with "good confidence marginal separation," 1 with a "questionable biopattern," and 6 "presumptive." The final accuracy was 92.3% for the Vitek GNI+. These results were similar to those reported by Bourbeau and Heiter (2), who had a 98.4% accuracy rate with 61 non-glucose-fermenting bacillus isolates.

In addition to Comamonas acidovorans for the W/A and Acinetobacter lwoffii and C. meningosepticum for the Vitek, Empedobacter brevis, P. fluorescens, Pseudomonas putida, P. stutzeri, and Sphingobacterium multivorum were the most common species requiring additional testing in both systems.

Tables 3 and 4 show the errors in the W/A and Vitek GNI+. Of the 12 errors in the W/A, 7 were misidentifications at the genus level, 3 were misidentifications at the species level, and 2 which were not in the database were interpreted as "very rare biotype." In this study, the most frequently misidentified species were S. paucimobilis and Agrobacterium radiobacter. Four of the six isolates of S. paucimobilis were misidentified as Sphingobacterium spiritivorum (three isolates) due to a false-negative reaction for maltose and as C. meningosepticum or C. indologenes (one isolate) due to a false-positive reaction for sucrose. Two of four isolates of A. radiobacter were both misidentified as Pseudomonas spp. due to a false-negative esculin hydrolysis reaction. These two species represented 3.3% (10 of 301) of all isolates tested in the W/A system.

Of the 23 errors in the Vitek GNI+ system, 12 were misidentified at the genus level and 4 were misidentified at the species level. The remaining seven isolates were interpreted as "unidentified." The most frequently misidentified species were A. lwoffii, C. indologenes, and C. acidovorans. Bourbeau and Heiter, O'Hara et al., and Pfaller et al. (2, 6, 7, 8) pointed out the weaknesses of the Vitek GNI and Vitek GNI+ systems in identifying A. lwoffii. For the GNI system, O'Hara et al. (7) reported that one of the A. lwoffii isolates was misidentified as Acinetobacter anitratus, and Pfaller et al. (8) reported that three out of five isolates of A. lwoffii were interpreted as "no identification." As for the GNI+ system, O'Hara et al. (6) reported one isolate and Bourbeau and Heiter (2) reported one out of three isolates were unable to be identified but were interpreted as no growth or "no identification" by using the GNI+ card. In our study, of the 10 isolates of A. lwoffii, two failed to grow in the GNI+ card, with an "unidentified" result, and two were misidentified due to false-positive citrate and/or urea reactions. The remaining 6 isolates were correctly identified after additional testing. Four of the 25 isolates of C. indologenes were misidentified as Brevundimonas vesicularis (one isolate) due to a false-negative polymyxin B reaction, as P. stutzeri (one isolate) due to a false-positive acetamide reaction, and as C. meningosepticum (two isolates), one due to a false-positive mannitol reaction and the other due to a false-negative glucose oxidation reaction. Three of the 12 isolates of C. acidovorans were misidentified. One was misidentified as Alcaligenes faecalis due to a false-positive polymyxin B reaction, and two were misidentified as Myroides odoratus due to false-negative mannitol reactions. These three species represented 15.6% (47 of 301) of all isolates tested in the Vitek GNI+ system.

In a study by O'Hara et al. (7), it took 34 min to enter data, generate bar codes, gather materials, and set up and load 10 panels for the W/A system compared to 21 min in our study. It took 30 min to gather material, make suspensions, and set up, fill, seal, and load 10 cards for the Vitek system, both in our study and in that by O'Hara et al. (7).

In summary, although they are both highly automated, additional testing was required for 24.6% of isolates with the W/A system and 20.6% for the Vitek GNI+ system. The additional tests take another 24 or 48 h, the same as reported by O'Hara et al. (7). In this study, both systems achieved an accuracy level above 90%. We concluded that there was no significant difference (P > 0.05) between these two systems.