Automated systems for identification of microorganisms.

Vol. 5, No. 3

CLINICAL MICROBIOLOGY REVIEWS, July 1992, p. 302-327

0893-8512/92/030302-26$02.00/0 Copyright X 1992, American Society for Microbiology

Automated Systems for Identification of Microorganisms CHARLES E. STAGER' AND JAMES R. DAVIS2* Department of Pathology, Ben Taub General Hospital, 1 and Department of Pathology, The Methodist Hospital,2 Baylor College of Medicine, Houston, Texas 77030

302 305 308 SENSITITRE ........................................................ WALKAWAY-96, WALKAWAY-40, AND AUTOSCAN-4 ........................................................ 310 315 ALADIN AND AUTOREADER ........................................................ 316 BIOLOG ......................................................... 318 MIDI MICROBLIL IDENTIFICATION SYSTEM ......................................................... 320 AUTOSCEPTOR ......................................................... STUDIES COMPARING AUTOMATED IDENTIFICATION SYSTEMS ..........................................321 321 DISCUSSION ........................................................ 324 ACKNOWLEDGMENTS ........................................................ REFERENCES .........................................................324 INTRODUCTION ........................................................ VITEK ........................................................

primary goal was to enhance data acquisition and processing, particularly with regard to decreasing turnaround time. Although the instruments available today are improvements over the original formulations, they still represent the first generation of instruments used to identify microorganisms. These instruments are widely accepted and very helpful; however, like the instruments used in clinical chemistry and hematology laboratories, they will continue to evolve to better meet the needs of the clinical microbiology laboratory. If we compare the modem clinical chemistry analyzer, with its discrete multianalytes requiring no sample preparation, with instruments available in clinical chemistry during the 1960s, we believe we get a glimpse of what the future can be in microbiology. At the very least, we should target that level of automation for clinical microbiology and expect future generations of equipment to be highly automated, cost-effective, accurate, reliable, and flexible and to provide rapid turnaround time. Among the first automated microbial identification systems were the Autobac Series II (formerly called the Autobac; Organon-Teknika, Durham, N.C.) and the Avantage Microbiology Center (formerly called the Abbott MS-2) and Quantum II Microbiology System (Abbott Laboratories Diagnostic Division, Irving, Tex.). These systems are no longer manufactured but are still in service in some laboratories. As an introduction to the systems used for automated identification, we believe that it is appropriate to provide a brief review of these systems. The Autobac Series II uses a 19-chambered plastic cuvette and automatically interprets results of biochemical tests. One chamber is a growth control, and the other 18 contain substrates composed of antibiotics, dyes, or other chemicals. Common members of the family Enterobacteriaceae and six species or groups of nonfermentative and oxidasepositive gram-negative bacilli can be identified by differential growth inhibition. After off-line incubation of the cuvette for 3 to 6 h, a photometer automatically determines growth inhibition by analyzing the light-scattering index of each chamber. A two-stage quadratic discriminant analysis program is then used to identify the isolate. Early studies of the Autobac Series II with currently used substrates demonstrated that 87.7 to 94.8% of the organisms

INTRODUCTION

Clinical microbiology has been an especially dynamic discipline during the past 10 to 15 years. The exciting developments include the recognition of several new etiologic agents, the reemergence of some classic pathogens, development of molecular diagnostic tools, and automation of antimicrobial susceptibility testing and microbial identification. This article explores the development of automated identification systems and reviews their performance. To limit confusion, we will avoid terms often used in the literature such as semiautomated or partially automated. Webster's Dictionary defines automation as the "automatically controlled operation of an apparatus, process, or system by mechanical or electronic devices that take the place of human organs of observation, effort, and decision" (77a). None of the systems described is totally automated. We use the term "automated" to describe the instruments discussed here and trust the reader to understand that some instruments are more automated than others. The criteria used for inclusion of an instrument in this review are as follows. (i) The minimum requirement is automated result entry and identification of microorganisms. Systems requiring manual result entry are not discussed. (ii) The instrument must have a data base for the identification of a large variety of different microorganisms. Instruments such as automated enzyme immunoassay systems that identify a relatively small number of microorganisms are not described. (iii) The instrument must be available in the United States. For studies that have compared the identification accuracy of two or more automated identification systems, the percentile (P value) of the chi-square distribution as determined by the chi-square test has been calculated. The development of the first generation of automated equipment for clinical microbiology involved essentially two approaches. One can be described as the mechanization of existing techniques. The second combined mechanization with other changes, such as miniaturization and/or incorporation of innovative substrates, inhibitors, or indicators. The *

Corresponding author. 302

VOL. 5, 1992

tested were correctly identified (10, 11, 18, 27, 60, 106, 111). Commonly isolated organisms such as Escherichia coli, Kiebsiella pneumoniae, Proteus mirabilis, and Pseudomonas aeruginosa were correctly identified. However, relatively common organisms such as Enterobacter cloacae, Citrobacter freundii, Proteus vulgans, Providencia spp., Salmonella spp., Shigella spp., and Pseudomonas cepacia, as well as uncommon organisms, were frequently incorrectly identified. In addition, when organisms not represented in the data base were tested, they were frequently given an identification. Only one of these early studies involved a comparison of the Autobac Series II with another automated identification system, the Vitek system (bioMerieux Vitek, Inc., Hazelwood, Mo.). The Vitek system will be described later in this paper. In this study, Barry et al. (10) tested 1,510 members of the family Enterobacteriaceae and nonfermentative gram-negative bacilli. The Autobac Series II correctly identified 1,443 (95.6%) of the isolates tested, and 1,457 (96.5%) were identified by the Vitek system (P > 0.10). Truant et al. (120) recently compared the Autobac Series II with the Vitek system and a visually read microtiter system. They tested 434 clinical isolates whose distribution was similar to that expected in the clinical setting. If the three systems disagreed on identification, the API 20E (Analytab Products, Plainview, N.Y.) or conventional biochemicals were used as the reference method. The Autobac Series II correctly identified 357 (82.3%) of the isolates. Organisms with a low relative-probability identification value or unidentified organisms accounted for 32% of the errors. Organisms that posed problems in this group included Citrobacter diversus, Enterobacter cloacae, E. coli, Proteus mirabilis, P. aenrginosa, and Serratia marcescens. Misidentification of the organism, particularly with Citrobacter spp., Enterobacter spp., and E. coli, accounted for 68% of the errors. The Autobac Series II also misidentified organisms from many different genera as either C. freundii or Enterobacter agglomerans. There was no discernible pattern to explain these misidentifications. The earlier studies of the Autobac also reported frequent problems in identification of Citrobacter and Enterobacter spp. but not E. coli, as reported by Truant et al. (120). In the study of Truant et al. (120), the Autobac Series II correctly identified 82.3% of the tested isolates, versus 95.6% for the Vitek system (P < 0.001). The Avantage Microbiology Center and Quantum II Microbiology System use a 20-chamber clear-plastic cartridge for identification of aerobic gram-negative bacilli or yeasts. The Abbott Bacterial Identification Cartridge (BIC) and Abbott Yeast Identification Cartridge (YIC) each contain 20 lyophilized substrates. The Avantage and Quantum II data base includes information for identification of 29 genera or species of the Enterobacteriaceae, 2 Acinetobacter spp., Xanthomonas maltophilia, 3 Pseudomonas species or groups, 5 other oxidase-positive gram-negative bacilli, 15 Candida spp., 8 Cryptococcus spp., S Rhodotorula spp., and 4 additional yeast genera. After inoculation of the cartridges, baseline turbidimetric or colorimetric readings are obtained for the biochemical chambers. With the Avantage, these optical readings are simultaneously determined by lightemitting diodes and matched photodetectors, whereas the Quantum II uses a dual-wavelength spectrophotometer to read biochemical chambers in sequence. After off-line incubation at 35 to 37°C for 4 to 6 h (BIC) or at 30°C for 22 to 24 h (YIC), readings are again obtained to determine turbidimetric or colorimetric changes. Isolates are identified by comparison of biochemical test results (oxidase and indole

AUTOMATED IDENTIFICATION SYSTEMS

303

test results for bacteria and the germ-tube test result for yeasts are manually entered) with a probability matrix. There have been two reports of the Avantage BIC with currently used substrates. Jorgensen et al. (55) reported the results of a collaborative evaluation of the Avantage BIC used to identify commonly isolated nonfermentative or oxidase-positive gram-negative bacilli in 5 h. Conventional biochemicals were used as the reference system. The organisms included in the data base were Acinetobacter anitratus, Acinetobacter iwoffi, Aeromonas hydrophila, Flavobacterium meningosepticum IIb group, P. aeruginosa, P. cepacia, Pseudomonas fluorescens-Pseudomonas putida group, X. maltophilia, and Plesiomonas shigelloides. In phase I of the study, 200 challenge strains were tested by each of three laboratories. The overall accuracy was 96%, with 95% correct results for isolates in the data base and 98% for recognition of biotypes not in the data base. Of 200 isolates, 11 either produced a correct identification with low likelihood (probability, 0.05). The Quantum II BIC correctly identified 111 (93.2%) of the nonenteric organisms tested, and the Vitek system identified 115 (96.6%) (P > 0.05). Rhoden and O'Hara (98) evaluated an updated Quantum II system, with changes in the software, an expanded data base, and the change of some biochemical formulations. The authors tested 335 isolates, which consisted of 258 members of the family Enterobacteriaceae, 55 nonfermenters, and 22 oxidase-positive fermenters. No more than 10 strains in each of 65 species were tested. The isolates consisted of both typical and atypical strains obtained from collections of the Centers for Disease Control and were not representative of those commonly found in the clinical laboratory. The isolates were identified by conventional biochemical and serologic meth-

304

STAGER AND DAVIS

ods. The Quantum II BIC correctly identified 92.6% of members of the family Enterobacteniaceae, 92.7% of the nonfermenters, and 91% of the oxidase-positive fermenters. To obtain these levels of correct identification, additional serologic (42 isolates), biochemical (34 isolates), or serologic and biochemical (17 isolates) tests were necessary as indicated by the Quantum II data base. The distribution of results was as follows: 184 (54.9%) were identified with a probability of 95% or greater; 51 (15.2%) were identified within the range of 80 to 94.9%; and 57 (17%) were identified with a probability of less than 80%. There were 25 misidentified organisms. Four nonfermenters (two P. fluorescens and two P. putida) and two oxidase-positive fermenters (Plesiomonas shigelloides) grew slowly, and this was believed to contribute to their misidentification. Three strains (one indole-negative E. coli, one Kluyvera cryocrescens, and one Yersinia enterocolitica) had biochemical profiles not recognized by the data base. The remaining 16 misidentified organisms were generally newly recognized genera, and the misidentifications were due to false-positive or false-negative test reactions. Cooper et al. (26) published a collaborative evaluation of the MS-2 YIC. In phase I of the study, 91% of 179 stock cultures of yeasts were correctly identified. Conventional methods were used as the reference system. Isolates difficult to identify included Trichosporon beigelii, Geotrichum spp., Candidafamata, and Candida humicola. In phase II, 96% of 378 clinical isolates were correctly identified. When there was a discrepancy between the routine laboratory test and the YIC, conventional methods were used for identification. Only 2% of those correctly identified had a low likelihood of correct identification (probability, 0.05). Recent abstracts regarding the Quantum II YIC have reported that 91 and 99% of the tested yeasts were correctly identified (5, 48). In summary, early studies with the Autobac Series II demonstrated that most of the gram-negative bacilli commonly isolated in the clinical laboratory were correctly identified (10, 11, 18, 27, 60, 106, 111). However, problems were reported in the identification of such organisms as Enterobacter cloacae, C. freundii, Proteus vulgaris, Providencia spp., Salmonella spp., Shigella spp., P. cepacia, and many uncommon organisms. Truant et al. (120) demonstrated additional unexplained problems in identification of some common organisms such as E. coli, Proteus mirabilis, and P. aeruginosa. The most recent versions of the Avantage BIC correctly identified approximately 95% of the gram-negative bacillus strains tested (55, 116), whereas the most recent version of the Quantum II BIC correctly identified approximately 92% of typical and atypical gramnegative bacillus strains tested (98). Organisms that particularly posed identification problems for the Abbott identification systems included P. fluorescens, P. putida, Serratia spp., and uncommon organisms. The MS-2 YIC (26), Avantage YIC (25), and Quantum II YIC (61, 94, 102, 104) performed well in the identification of common yeast isolates, but less commonly encountered yeasts were frequently misidentified. The number of genera, species, or groups of gram-negative bacilli or yeasts in the data base of the Autobac Series II or Abbott identification systems is limited. Consequently, these early identification systems have had significant problems in identification of uncommon organisms and newly recognized genera. These first automated identification systems are of historical importance and can serve as a reference of performance level for the systems to be described in this review.

VOL. 5, 1992


VITEK

The Vitek system had its origins in the 1960s, when McDonnell Douglas was contracted by the National Aeronautics and Space Administration to develop an automated system to detect and identify pathogens directly from urine specimens of astronauts in space. This system was subsequently modified and introduced to the clinical microbiology laboratory as the AutoMicrobic System (AMS) in 1976. The Vitek system, based on bacterial growth in microwells of thin plastic cards, could identify nine common urinary tract pathogens directly from urine specimens and used the mostprobable-number concept to determine the presence of more than 4.5 x 104 CFU per ml of microbes of urine. Vitek later developed additional cards that required pure cultures for inocula. These 30 microwell cards contained antibiotics or biochemical substrates. Susceptibility test cards are available for both gram-negative bacilli and gram-positive bacteria with 11 antimicrobial agents per card. Results are available in 4 to 8 h. Vitek has a variety of standard test kits, and custom-defined test kits can be purchased. Over 40 antimicrobial agents are currently available on the cards, and results from each test include an interpolated MIC, as well as the National Committee for Clinical Laboratory Standards categories of susceptible, moderately susceptible, intermediate, and resistant. Identification cards automatically interpreted by the Vitek system are the Gram-Negative Identification Test Kit (GNI), the Gram-Positive Identification Test Kit (GPI), and the Yeast Biochemical Test Kit (YBC). Identification cards that require off-line incubation and manual entry of the results into the Vitek computer are the Anaerobe Identification Test Kit, Neisseria/Haemophilus Identification Test Kit, and Enteric Pathogen Screen Test Kit. The GNI and GPI each contain 29 substrates and a growth control medium. The GNI substrates include 25 conventional biochemical substrates, 3 proprietary substrates, and 1 antibiotic. The GNI must be marked if the organism is oxidase positive. The GNI data base includes information for identification of 46 species of members of the family Enterobactenaceae and 39 species of other gram-negative organisms. The GPI substrates include 26 based on conventional biochemical tests, two antibiotics, and one dye. The GPI must be marked for catalase-negative, beta-hemolytic organisms or for coagulase-positive organisms that are catalase positive. The GPI data base includes information for identification of 23 Streptococcus species, 4 Enterococcus species, 16 Staphylococcus species, and 4 Corynebacterium, Aerococcus, Listenia monocytogenes, and Erysipelothrix rhusiopathiae species or groups. The YBC contains 26 substrates which are based on conventional methods. The YBC data base includes information for identification of 16 Candida species, 6 Cryptococcus species, 3 Rhodotorula species, 2 Tnichosporon species, 3 Geotnichum species, 2 Prototheca species, and single species of four additional genera. The Vitek system is an integrated modular system consisting of a filling-sealer unit, reader-incubator, computer control module, data terminal, and multicopy printer. The Vitek system can be purchased with a capacity of 30, 60, 120, or 240 cards and can be interfaced with other computers. A data management center can be added. Inocula for the identification cards are prepared from selective (GNI or GPI) or nonselective (GNI, GPI, and YBC) agar media. Inocula for the GNI, GPI, and YBC are prepared by suspending several colonies in 1.8 ml of 0.45 to

305

0.5% saline and adjusting the suspension to the equivalent of a no. 1 (GNI and GPI) or a no. 2 (YBC) McFarland standard. The inoculum is automatically transferred to the card via a transfer tube during the vacuum cycle of the filling module. The GNI and GPI are placed in plastic trays, each tray holding up to 30 cards. The tray is placed in the readerincubator at 35°C, and at hourly intervals a digitized analog optical reading, proportional to the light attenuation for each test well, is obtained for each card. The first reading usually establishes a baseline value, and the amount of light reduction caused by growth or a biochemical reaction in the microwell is determined on subsequent readings. A predetermined minimum change is required to differentiate between positive and negative reactions. Final identification by the GNI is reported between 4 and 18 h. Most of the non-glucose-fermenting gram-negative bacilli are reported at 18 h. Organisms are identified by the GPI between 4 and 15 h. The YBC is incubated off-line at 30°C for 24 h and then placed in the reader-incubator for a single reading. A message "reincubate for 24 h at 30°C" indicates that a definitive identification requires more incubation time. At 48 h, one must fill in the 48H mark on the card and obtain a second reading. The biochemical test results for all cards are compared with the data base, and the first and second choices, as well as their absolute likelihoods and normalized percent probabilities, are reported. The biochemical test results, as well as supplemental tests if required, are printed. The Vitek system original gram-negative identification card was called the Enterobacteriaceae Biochemical Card (EBC) and was designed for the identification of members of the family Enterobacteriaceae within 8 h. The EBC contained 23 conventional biochemical substrates, 2 inhibitors, and 1 nonconventional substrate. Even though the original Vitek system had a limited data base, numerous studies found the EBC to correctly identify 92 to 99% of the tested organisms (7, 12, 17, 29, 39, 41, 50, 53, 58). Organisms not represented in the data base, uncommon organisms, and incorrect reactions for certain biochemicals such as adonitol, arginine, citrate, H2S, and malonate were responsible for most misidentifications. The lack of an indole reaction was also considered a weakness of the system. Ferraro et al. (39) found that commonly isolated members of the family Enterobacteriaceae could be presumptively identified by the EBC in 4 h. They found that 97% of the members of the Enterobacteriaceae isolated in their laboratory belonged to 11 species of six genera. When the EBC was limited to those 11 species, 83% of the isolates were correctly identified to genus or species at 4 h. Two percent of the isolates were misidentified, and 15% were not reported until 8 h. Vitek then introduced the EBC+, which, with the addition of acetamide, cetrimide, glucose oxidation, and a supplemental oxidase test, was capable of identifying members of the family Enterobacteriaceae within 8 h and nonfermentative and oxidase-positive gram-negative bacilli within 18 h. Studies of the EBC+ demonstrated that 90 to 96% of the members of the family Enterobacteriaceae and 86 to 97% of the gram-negative organisms that were not members of the Enterobacteriaceae tested were correctly identified (7-10, 44, 56, 125). However, uncommon organisms were frequently misidentified or had a low percent likelihood of identification. The new substrates added to the EBC+ were of limited value in the identification of gram-negative organisms that were not members of the Enterobacteriaceae, and only Aeromonas iwoffi, Aeromonas hydrophila, and P. aeruginosa were consistently correctly identified (7, 8, 54, 108, 125). Barry and Badal (8) determined the reliability of

306

STAGER AND DAVIS

the EBC+ for early preliminary identification of isolates. They reported that Vitek system identifications were 96% accurate at 4 h if all results with a probability of 80%) was obtained for most S. haemolyticus, S. simulans, Staphylococcus capitis, Staphylococcus cohnii, and S. xylosus isolates, but only 35.2% of Staphylococcus hominis, 69.5% of Staphylococcus warnei, and 60% of Staphylococcus sciun isolates were properly identified. Supplemental tests were not required with the GPB. The only reading errors by the autoSCAN-4, compared with visual readings, were observed with bacitracin, crystal violet, or novobiocin susceptibility and were due to air bubbles or scratches on the plastic well. The GPB biochemical tests that differed from conventional tests included the pyrrolidonyl3-naphthylamide, ,-D-galactopyranosidase, lactose, and urease tests (false-negative reactions) and the arginine, mannose, novobiocin, and lactose tests (false-positive reac-

tions). Kloos and George (62) compared the GPB and R-GPB tests for the identification of 25 different Staphylococcus spp. The 920 strains tested were obtained from reference and type collections, clinical specimens, and the skin of humans and animals. The nonreference strains were identified by the characteristics and methods described by Kloos and Lambe (63), and selected strains from each species were verified via DNA-DNA hybridization with reference or type strains. The GPB was read visually at 15 to 48 h, and the R-GPB was read by the WalkAway-96 at 2 h. Both systems correctly identified 290% of strains of Staphylococcus arlettae, S. aureus, Staphylococcus auricularis, S. capitis subsp. capitis, Staphylococcus camosus, S. cohnii subsp. cohnii, Staphylococcus lentus, S. saprophyticus, and S. sciuri. In addition, the R-GPB correctly identified >90% of strains of Staphylococcus caprae, Staphylococcus caseolyticus, S. epidermidis, Staphylococcus kloosii, S. warneri subsp. 2, and S. xylosus. Both systems correctly identified between 80 and 90% of strains of Staphylococcus equorum, S. haemolyticus subsp. 1, Staphylococcus hyicus, and Staphylococcus intermedius. The R-GPB also correctly identified between 80 and 89% of strains of S. capitis subsp. ureolyticus, S. cohnii subsp. urealyticum, S. hominis, S. simulans, and S. wameri subsp. 1. Both systems identified 50 to 75% of strains of S.

chromogenes, Staphylococcus gallinarum, Staphylococcus lugdunensis, and Staphylococcus schleiferi. Stoakes et al. (114) evaluated the WalkAway-96 R-GPB

with 239 strains of staphylococci belonging to 17 species. Ten or more strains of commonly occurring clinical strains were tested. Conventional methods were used to establish the true identity of the strains. Of the tested strains, 219 (91.6%) were correctly identified without additional tests, 9 (3.8%) were correctly identified after additional tests, 6 (2.5%) were incorrectly identified, and 5 (2.1%) were classified as rare biotypes and not identified. The misidentified stains (all considered common isolates), were S. capitis (2 of 21 strains), S. hominis (2 of 20 strains), S. wameri (1 of 25 strains), and S. xylosus (1 of 12 strains). One discrepant reaction (positive in each instance) was responsible for each misidentification. Of the 37 (representing 7 species) less


313

commonly isolated Staphylococcus strains, none were incorrectly identified and only 1 (Staphylococcus lugdunensis) required additional tests for correct identification. Godsey et al. (43) evaluated the WalkAway-96 R-GPS for identification of 607 staphylococci, 704 streptococci, 42 Micrococcus isolates, 28 Aerococcus isolates, and 30 L. monocytogenes isolates. The accuracy of identification ranged from 79% (S. warnen) to 100% (multiple species). The overall identification accuracy was 95.7%. Stoakes et al. (113) evaluated the AIP both by visual interpretation and with the autoSCAN-4. The results were compared with those obtained by the Virginia Polytechnic Institute conventional method. A total of 237 anaerobes were tested, of which 166 (70%) were correctly identified to species level (probability .85%) by visual readings versus 157 (66.2%) correctly identified by the autoSCAN-4 (P > 0.30). When supplemental tests were performed, 80.1 and 76.7%, respectively, were correctly identified (P > 0.30). The visual and automated interpretations produced 94 discrepant reactions, including 37 with gram-negative bacilli, 48 with clostridia, and 9 with other organisms. The auto SCAN-4 could not interpret a reaction in 46 instances, whereas 93 reactions were difficult to interpret visually. Difficulties in interpretation of reactions were randomly distributed with the autoSCAN-4 but were frequent for

L-lysine-,3-naphthylamide reactions with Bacteroidesfragilis and for bis-p-nitrophenyl phosphate reactions with grampositive organisms when read visually. To determine the reproducibility of the system, 50 strains (30 gram-negative bacilli and 20 clostridia) representing 1,200 reactions were tested in duplicate. Discrepant results were 69 (5.8%) with the autoSCAN-4 and 57 (4.8%) with visual readings (P > 0.20). As a result, 8% of the strains were changed from a correct to an incorrect identification with the autoSCAN-4 and 6% with visual readings. Land et al. (67) evaluated the YIP with 437 recent clinical yeast isolates, using the API 20C as the reference method. However, the MicroScan automated instruments were not used and only visual readings were obtained. The YIP identified 85% of all isolates and 92% of the taxa included in the data base. The YIP identified 94% of rapidly growing yeasts in the genera Candida, Hansenula, Pichia, Rhodotorula, Saccharomyces, and Torulopsis (98% when organisms and biocodes not in the data base were excluded). However, only 65% of identifications of slower-growing yeasts (Blastoschizomyces, Cryptococcus, Geotnichum, Hyphopichia, Phaeococcomyces, Prototheca, and Trichosporon species) correlated with results of the API 20C. When unrecognized organisms and biocodes were excluded, the correlation improved only to 68%. The yeasts that posed the greatest problem for the YIP were Torulopsis beigelii, Blastoschizomyces capitatus, Geotrichum spp., and Cryptococcus neoformans. The YIP correctly identified 40 (67%) of 60 known serotypes of Cryptococcus neofornans. One strain of serotype A was incorrectly identified, and 19 isolates (3 of 14 serotype A, 3 of 15 serotype B, 2 of 15 serotype C, and 11 of 15 serotype D isolates) yielded biocodes not found in the data base. The authors concluded that the expansion of the YIP data base and adjustment of the indoxyl phosphatases and pH indicators for sucroses 1 and 2 and trehalose should make the YIP an excellent system for identification of medically important yeasts.

St.-Germain and Beauchesne (112) reported on a modified version of the MicroScan yeast identification system. Four substrates (isoleucine, urea, N-acetylgalactosamine, and trehalose) had been reformulated and the data base regener-

314

STAGER AND DAVIS

ated for Version 17 software. The authors evaluated the YIP for identification of 357 yeastlike clinical isolates of 11 genera and 30 species. Of these, 217 were common isolates and 140 were relatively uncommon. The YIP results were read both visually, by two observers, and with the auto SCAN-4. The API 20C and morphological characterization on cornmeal-Tween 80 agar were the reference systems. Conventional tests were used to resolve any discrepancies between the API 20C and the YIP. Both the YIP, as interpreted by the autoSCAN-4, and the API 20C identified 278 (78%) of 357 strains without supplemental tests. With supplemental tests, the YIP correctly identified 345 strains, incorrectly identified 10 strains, and did not identify 2 strains, for an overall accuracy of 96.6%. It accurately identified 99.5% of the common strains and 92.1% of the less common strains. Of the common strains, the autoSCAN-4 incorrectly identified only one strain of C. glabrata (99.5% accuracy), whereas none were incorrectly identified when the panels were visually interpreted. Of the less common strains, the autoSCAN-4 incorrectly identified nine strains and failed to identify two strains (92.1% accuracy). On visual interpretation 12 strains were incorrectly identified and 2 or 3 strains were not identified (89.3% accuracy). There was an overall accuracy of 96.6% for the autoSCAN-4 versus 95.8% for visually interpreted panels (P > 0.50). Identical profile numbers were obtained by both observers and the auto SCAN-4 for 44% of the 357 strains. Difficulties with the interpretation of chromogenic substrates, particularly ,-naphthylamide substrates and nitrophenyl-linked substrates, resulted in discrepancies in profile numbers between the two observers and the autoSCAN-4. Abstracts regarding the YIP have recently been published. Simmonds et al. (107) reported on the YIP for identification of 232 clinical yeast isolates from 10 genera and 29 species. The API 20C was used as the standard method, and discrepancies were resolved with conventional assimilation and fermentation tests. The overall accuracy of the YIP was 59% when interpreted by the autoSCAN-4 and 72% when interpreted visually (P < 0.01). When biotypes not found in the data base were excluded, the accuracy was 62% by auto SCAN-4 and 76% by visual interpretation (P < 0.01). Morphological observations were needed to identify more than 50% of the isolates, and confirmatory biochemical tests were required for identification of 29% of the isolates by the autoSCAN-4 versus 16% by visual interpretation. Belcher et al. (14) tested the YIP with 189 clinical yeast isolates from 7 genera and 20 species. The Minitek Yeast Carbon Assimilation procedure (Becton-Dickinson) was the standard for comparison. The autoSCAN-4 correctly identified 79% of the isolates, whereas visual interpretation correctly identified 86% (P > 0.05). No supplemental morphological or physiologic tests were performed. Stockman and Roberts (115) tested the YIP with 260 yeast isolates from 24 species. The API 20C was the reference method. It was not stated whether readings were obtained by an automated system or by visual inspection. The YIP correctly identified 75.3% of the isolates with a probability of >85%. Supplemental tests were required for 21% of the isolates. Two C. famata, two Candida zeylanoides, and one each of C. glabrata, C. lipolytica, C. parapsilosis, C. tropicalis, Kluveromyces lactis, S. cerevisiae, and Sporobolomyces isolates were misidentified by the YIP. In summary, the most recent study of the autoSCAN-4 GNB correctly identified 95.4% of tested members of the family Enterobacteriaceae to the species level (42). Pfaller et al. (95) reported that the WalkAway-96 R-GNB correctly

CLIN. MICROBIOL. REV.

identified 86.3% of members of the family Enterobacteriaceae and 92.3% of gram-negative nonenteric bacilli tested. Supplemental tests were required for 7.6% of the isolates for correct identification. Tenover et al. (117) found that the WalkAway-96 GNB and R-GNB correctly identified 62.6 and 90.6%, respectively, of the gram-negative nonenteric bacilli tested. However, approximately 35 and 45%, respectively, of the correct identifications with the GNB and R-GNB had a probability of less than 85%. In addition, approximately 40% of the isolates required supplemental tests for correct identification by both the GNB and R-GNB. Only 13.5% of the GNB results could be interpreted at 18 h, with the remainder requiring 42 h of incubation. A. xylosoxidans subsp. xylosoxidans, P. putida, P. fluorescens, and X. maltophilia were most frequently misidentified by the GNB, whereas P. fluorescens accounted for 28% of the misidentifications by the R-GNB. Published abstracts concerning the WalkAway-96 R-GNB have reported an identification accuracy that has ranged from 66.4 to 97.8% (24, 30, 32, 47, 57, 59, 70, 75, 76, 87, 118). There have been no studies of the autoSCAN-4 involving the GPB panel with dried substrates. Kloos and George (62) found that the WalkAway-96 R-GPB correctly identified common Staphylococcus spp. and many uncommon Staphylococcus spp. at a probability of >85%. Stoakes et al. (114) found that the WalkAway-96 R-GPB correctly identified 91.6% of Staphylococcus spp. tested without supplemental tests and 95.4% with required supplemental tests. None of the less commonly isolated Staphylococcus spp. (37 of 239 strains) were incorrectly identified. Godsey et al. (43) tested 1,411 gram-positive bacteria, including staphylococci, streptococci, Micrococcus isolates. Aerococcus isolates, and L. monocytogenes, with the WalkAway-96 R-GPB and found an overall identification accuracy of 95.7%. The only evaluation of the AIP was with the autoSCAN-4 (113). When required supplemental tests were performed, the autoSCAN-4 AIP correctly identified 76.7% of the anaerobes tested (N = 237). Results for correct identification with the autoSCAN-4 compared favorably with visual interpretation of the same panels (P > 0.30). St.-Germain and Beauchesne (112) reported on a recent modified version of the autoSCAN-4 YIP and found that 99.5% of the common yeast isolates and 92.1% of the less common yeast isolates tested were correctly identified when the required supplemental tests were performed. There was an overall accuracy of 96.6% for the autoSCAN-4 versus 95.8% when the same panels were visually interpreted (P > 0.50). Recent abstracts on the autoSCAN-4 YIP have reported that 62 and 79% of the yeasts tested were correctly identified (14, 107). There have been no published studies of the WalkAway-96 YIP. The MicroScan automated systems will identify a wide variety of organisms. However, further improvements in the data base and software for all panel types are needed. An automated method for the addition of mineral oil to the various panels and for inoculation of the HNID, AIP, and YIP would enhance work flow. There have been no evaluations of the HNID panel and GPB dried-substrate panels by using MicroScans automated systems. Evaluation of the WalkAway-40 and further evaluation of the WalkAway-96 and autoSCAN-4 with appropriate MicroScan panel types will be required as Micro Scan continues to update the data bases and software programs.

VOL. 5, 1992


ALADIN AND AUTOREADER The Automated Laboratory Diagnostic Instrument (ALADIN; Analytab Products) is a computer-assisted system with on-line incubation that interprets biochemical and susceptibility tests by a video image process. Analytab Products also manufactures the UniScept system AutoReader, which tests the same panels as the ALADIN but requires off-line incubation and which performs automatic interpretation of results through photometric evaluation of test wells at multiple wavelengths. The ALADIN and UniScept systems perform both MIC and qualitative susceptibility tests on aerobic gram-negative bacilli and gram-positive bacteria. Both systems identify gram-negative bacteria (UniScept 20E), grampositive bacteria (UniScept 20GP), and anaerobes (ANIdent). Other Analytab Products identification panels can be used with either system, but manual entry of the reactions is required. These panels identify gram-negative bacteria (Rapid-E and Rapid NFT), Staphylococcus spp. (Staph-Ident), and yeasts (Yeast-Ident). The UniScept 20E has 20 modified conventional substrates, and the data base includes information for identification of 55 species of the family Enterobacteriaceae and 50 groups, genera, or species of nonfermentative, oxidasepositive, gram-negative bacilli in 18 to 24 h. The UniScept 20GP has four chromogenic, one ,B-naphthol-labeled, one 0-naphthylamide-labeled, and 14 modified conventional substrates, and the data base includes information for identification of 13 Staphylococcus spp., 3 Enterococcus spp., and 2 Streptococcus spp. (group D, nonenterococcus) in 18 to 24 h. The AN-Ident has nine chromogenic substrates whose reactions are detected by the liberation of indoxyl, o-nitrophenol, or p-nitrophenol from the substrates, nine 3-naphthylamide-labeled substrates, one modified conventional substrate, and a test for catalase. The AN-Ident data base includes information for identification of 83 groups, genera, or species of anaerobic bacteria in 4 h. The ALADIN system is composed of an incubator, elevator, reagent-dispensing station, grabber arms, image processor, and disposal station. A separate workstation interacts with the ALADIN and is composed of a keyboard, computer (Compaq 386 Sx model 40), UniScept deziner-er Software, and Okidata 320 printer. The software contains the data base for all UniScept products, susceptibility programs, patient demographics, storage of data, and antibiograms. A bidirectional mainframe computer interface that allows up-load of daily test results and down-load of demographics may be purchased. Inocula for the UniScept 20E and UniScept 20GP can be grown on either selective or nonselective agar media, whereas inocula for the AN-Ident must be grown on nonselective agar media (excluding 5% sheep blood Trypticase soy agar [TSA] or Schaedler's blood agar). To prepare inocula, several morphologically similar colonies are selected and suspended in sterile 0.85% saline (UniScept 20E and Uni Scept 20GP) or sterile distilled water (UniScept 20E, Uni Scept 20GP, and AN-Ident). The suspension is adjusted to the equivalent of a no. 0.5 McFarland turbidity standard (UniScept 20E and UniScept 20GP) or a no. 5 McFarland standard (AN-Ident). The inocula for the UniScept 20E and UniScept 20GP are diluted 1:100 in sterile 0.85% saline. With either the UniScept Autoinoculator or the manual pipetting device, 100 ,ul of inoculum is dispensed into each microtube of the UniScept 20E (300 RI into the citrate, VogesProskauer, and gelatin microtubes) and UniScept 20GP. For the AN-Ident, each microtube is manually filled with 2 drops

315

(approximately 85 i±l) of the suspension. When the UniScept 20E is to be incubated off-line and then read on the UniScept AutoReader, the cupule section of the arginine, lysine, ornithine, urea, and H2S tubes must be manually overlaid with mineral oil. The ALADIN has a 60-specimen-capacity incubator, with up to two UniScept panels (identification and/or susceptibility) being processed per specimen. This allows the processing of up to 120 individual panels. A virtually unlimited number of panels may be incubated off-line for reading on the ALADIN. One or two UniScept panels per specimen are placed in a universal carrier, which is a molded plastic frame approximately 5 by 85/8 in. (ca. 13 by 22 cm) in size. The panels are inoculated, and the universal carrier is placed into 1 of 60 test slots (two rows of 30) in the incubator. The elevator and grabber arms automatically transfer the panels to the read station. The specimen number and panel type are interpreted by video image processing. This activates the appropriate incubation cycle and reagent addition for each panel. After appropriate incubation, panels are transferred to the reagent-dispensing station for addition of reagents; after further appropriate incubation, they are returned to the reader for a final examination. Currently, up to nine reagents can be dispensed. Each microtube is examined by a blackand-white video image processor through one to four colored filters selected by the computer, depending on which reaction is being analyzed. The video imaging camera isolates an area of interest for each microtube and uses 200 to 300 picture elements, or pixels, to view this area. For example, fermentation reactions are read at the middle or bottom of the microtube. The density of each pixel perceived by the camera is a specific voltage that is represented digitally for computer processing. These specific results are converted into plus or minus reactions, and the computer generates a seven-digit profile number. This profile number, or "biotype," identifies the most probable organisms of each biotype. The panels are then transferred to the disposal station, where they are discarded into a waste bag by miniature forklifts. Panels incubated off-line are placed in a tray and read on the UniScept AutoReader or the ALADIN. For the Uni Scept AutoReader, a photometer examines each microwell at multiple positions and wavelengths. A photodiode detects the transmitted light, and the resulting voltage is converted to an optical density value, which is equated to a predetermined positive or negative reaction. The resulting biochemical profile is compared with the data base for identification. Shulman et al. (105) reported on the ability of the ALADIN to read UniScept 20E panels. A total of 300 isolates (144 E. coli, 45 Proteus/Providencia spp., 27 KiebsiellalEnterobacter spp., 18 Serratia spp., 12 Pseudomonas spp., and 54 other organisms) were tested. On a test-for-test basis, the ALADIN and visual readings of 300 organisms showed a 99.0% level of correlation. These differences in readings did not affect the final identification of any of the tested organisms. The reference system to establish the true identity of the organisms and the percentage of organisms correctly identified was not mentioned. Navarro et al. (86) reported on the ALADIN for its ability to read AN-Ident panels. A total of 125 anaerobes (48 Clostridium spp., 37 Bacteroides spp., 15 Fusobacterium spp., 9Actinomyces spp., 5 Propionibacterium spp., and 11 other organisms) were tested. With respect to visual readings of the 125 organisms, the ALADIN showed a 96.1% level of correlation on a test-for-test basis. The abstract did not mention whether these differences in readings affected

316


STAGER AND DAVIS

the final identification of any of the tested organisms by the ALADIN. Also, the reference system to establish the true identity of the organisms and the percentage of organisms correctly identified were not mentioned. The only extensive report on the ALADIN identification system was a collaborative study at three medical centers (28). In this study, 318 aerobic and facultatively anaerobic gram-negative bacteria and 148 obligately anaerobic isolates were tested. All bacteria were identified by established Lonventional methods. The UniScept 20E and AN-Ident biochemical reactions were interpreted on the ALADIN by video imaging. Then the panels were visually interpreted. The results were compared on a test-for-test basis, with the visual interpretation serving as the reference. When two or more discrepant biochemical test results per test system were obtained, the isolate was retested. The results of the retest were considered final. Altogether, 6,360 individual tests were performed on the 318 gram-negative bacteria, with an overall agreement of 96.5%. False-negative readings for indole production, mannitol fermentation, and tryptophan deaminase produced =90% agreement. The authors suggested that the computer algorithms for these substrates, all carbohydrates, and the o-nitrophenyl-3-D-galactoside reaction be adjusted. With anaerobic test results, the overall agreement between the ALADIN and visual interpretation was 95.2%. False-positive results with indoxylacetate yielded an 81.1% agreement for this substrate. False-negative results, particularly with carbohydrate and some ,B-naphthylamide derivative tests, were noted. It was suggested that the indole test might be improved by the substitution of dimethylaminocinnamaldehyde for p-dimethylaminobenzaldehyde. This has now been incorporated into the ALADIN. Intralaboratory and interlaboratory reproducibility with the ALADIN biochemical profiles averaged 96.0 and 91.5%, respectively, in the three laboratories. Similar values were obtained for visually determined results. The report of D'Amato et al. (28) verified that there was good agreement between video image and visual interpretation of biochemical reactions for the UniScept 20E and AN-Ident, but it did not report the effect of false-positive or false-negative results by video image on identification of the organisms tested. There are some limitations to the use of the UniScept 20E and the ALADIN. With the UniScept 20E, some slowgrowing members of the family Enterobacteriaceae and gram-negative organisms that are not members of the Enterobacteriaceae require supplemental tests (oxidase, oxidation-fermentation glucose medium, or motility medium) and incubation for 36 to 48 h. The ALADIN will not identify these organisms, since reagents are added and panels are read at 24 h. These panels are automatically returned to the incubator for removal and off-line incubation. The only study of the UniScept system AutoReader was reported by O'Hara et al. (90). They compared the agreement of automated and visual readings of biochemical tests with the results of the UniScept 20E. They tested 291 oxidase-negative and 49 oxidase-positive glucose-fermenting isolates. Since the AutoReader is not designed to read a UniScept 20E at 48 h, nonfermentative bacteria were not evaluated. Of the 6,800 tests compared, only 45 readings (0.7%) did not match. Discrepant results involved 16 biochemicals, with indole and citrate disagreeing most often (13 of 45). The ALADIN and UniScept system AutoReader identify a limited number of gram-positive bacteria. Although the

organisms that can be identified by the UniScept 20GP represent most of the gram-positive bacteria routinely encountered in the clinical laboratory, it is desirable to have an automated system that could identify a broader range of significant gram-positive bacteria. The only rapid (4-h) panel that can be automatically read by the ALADIN and Uni Scept system AutoReader is the AN-Ident. Automated reading of other rapid Analytab Products identification panels would enhance the usefulness of these systems. It is commonly accepted that the UniScept 20E, UniScept 20GP, and AN-Ident have excellent data bases for identification of bacteria. However, there are no reported studies with either the ALADIN or UniScept system AutoReader and the UniScept 20GP. The studies by D'Amato et al. (28) and O'Hara et al. (90) indirectly suggest that the ALADIN and UniScept system AutoReader, respectively, will reliably identify isolates with either the UniScept 20E or the ANIdent. However, studies to prove the accuracy of identification of the UniScept 20E, UniScept 20GP, and AN-Ident when tested on these systems have not been reported. BIOLOG The Biolog system (Biolog, Inc., Hayward, Calif.) was introduced in 1989 for identification of aerobic gram-negative bacteria (enteric bacilli, nonfermenters, and fastidious species) by determination of carbon source utilization profiles. Recently, Biolog has added the capability to identify a broad range (cocci, bacilli, and spore-forming bacilli) of aerobic gram-positive bacteria. The Biolog GN MicroPlate (for identification of gram-negative bacteria) and the Biolog GP MicroPlate (for identification of gram-positive bacteria) are 96-well dehydrated panels containing tetrazolium violet, a buffered nutrient medium, and a different carbon source for each well except the control, which does not contain a carbon source. The microwells are rehydrated with a cell suspension and read at either 4 h or overnight (16 to 24 h) for the ability of the bacteria to utilize the carbon source. Tetrazolium violet is a redox dye used to detect electrons donated by NADH to the electron transport system. Reduced tetrazolium violet is a purple formazan. When a carbon source is not used, the microwell remains colorless, as does the control well. The resulting pattern of purple wells yields a "metabolic fingerprint" of the bacterium tested (15,

16). For metabolic capability studies, the MT MicroPlate contains only tetrazolium and a buffered nutrient medium without a carbon source in any of the 96 wells. The user can add various carbon sources to the microplate. The ES MicroPlate contains 95 different carbon sources and is designed for characterizing and/or identifying different E. coli and Salmonella strains, for mutant strain characterization or for quality control testing of E. coli K-12 or Salmonella typhimurium LT2 strains carrying recombinant plasmids, and for epidemiological studies. Although there is no Biolog data base for the MT MicroPlate and ES MicroPlate, a custom data base can be created by using Biolog software. Biolog is developing the capacity for identification of yeasts, anaerobes, and additional environmental and oligotrophic bacteria. The 95 substrates contained in the Biolog GN MicroPlate and the Biolog GP MicroPlate include carbohydrates, carboxylic acids, amides, esters, amino acids, peptides, amines, alcohols, aromatic chemicals, halogenated chemicals, phosphorus- and sulfur-containing chemicals, and polymeric chemicals. The Biolog system data base includes informa-


VOL. 5, 1992

tion for identification of 569 species or groups of aerobic gram-negative bacteria and 225 species or groups of grampositive bacteria and encompasses almost all known human pathogens and most important environmental species. The Biolog automated system consists of a manual eightchannel repeating pipettor, a turbidimeter, a MicroPlate Reader, a MicroLog Program Disk, and any DOS-based IBM-compatible PC, including XT and AT 286, 386, and 486 models. A manual system without the MicroPlate Reader is also available. The computer must have a hard drive of at least 20 MB, at least one floppy drive (3.5 or 5.25 in. [ca. 8.9 or 13.3 cm]), and, when the automated reader is used, an enhanced 101 keyboard. The software runs on systems with VGA or EGA color graphics or with monochrome displays. Different software versions can be purchased, either for manual entry or for automated reading of the test results. The latter software version allows data to be printed and saved in computer files, which can be utilized by userprovided software routines. Alternatively, the data can be filed in a "user data base" in which the reaction patterns are permanently saved. An unknown biochemical profile can be compared with the Biolog GN or Biolog GP data base, the user data base, or a combination of the two. Other features of the software include on-line information about any species in the library, cluster analysis programs in the form of dendrograms and two-dimensional and three-dimensional plots to demonstrate the relatedness of strains or species, and the separation of the GN data base into clinical and environmental versions. The inoculum for gram-negative organisms is prepared from TSA or TSA-blood agar medium. The inoculum for gram-positive clinical and food isolates, with few exceptions, must be grown on Biolog Universal Growth Medium with 5% sheep's blood (Biolog, Inc.). Environmental isolates typically do not require the addition of blood. Generally, bacteria are grown for 4 to 18 h. A swab is gently rolled over the colonies to prevent carryover of nutrients from the agar medium into the saline (0.85%) suspension of bacteria. The colorimeter or spectrophotometer (optical density at 590 nm of 0) is blanked with a tube containing uninoculated saline. The bacterial suspension is adjusted to within a low-standard-to-high-standard range. The inoculum should be used within 10 min. The plates are inoculated with 150 ,ul per well and incubated at 30 to 35°C either with or without CO2. The plates are read at 4 h either manually or on a computercontrolled microplate reader. The MicroLog software subtracts the background cell density from the negative control well and interprets all tests above a threshold as positive. The metabolic profile of the organism is matched to a data base of patterns by using the MicroLog Software Program. Data from the MicroPlates can be permanently saved in computer files to facilitate subsequent analyses. The automated reader processes a plate in 5 s. There have been four published studies on the use of the Biolog GN, one of which used the MicroPlate Reader. Carnahan et al. (19) tested 20 clinical strains each of Aeromonas hydrophila, Aeromonas caviae, and Aeromonas sobria with the Biolog GN. The reference methods for identification of the isolates were not stated. Isolates were grown overnight on TSA, and the Biolog GN was inoculated as specified by the manufacturer. Inoculated plates were incubated at 35°C for 18 to 20 h and then read manually. For the three species tested, nine substrates were found to yield good, discriminatory values. Seven of these substrates had not been previously identified as being useful for identifying Aeromonas isolates to the species level. All 60 Aeromonas

317

strains were correctly classified to species. However, Aerospecies were not included in the Biolog data base at that time, and this study demonstrated only that Aeromonas isolates could be identified to species with substrates present in the Biolog GN panel. Armon et al. (4) tested the Biolog GN panel with Legionella spp., also before this organism was included in the Biolog data base. They tested one strain each of Legionella

monas

pneumophila (Philadelphia 1), Legionella pneumophila (environmental isolate), Legionella micdadei, Legionella oakridgensis, Legionella longbeachae, and Legionella gormanii. The strains were identified by reference biochemicals but were not serogrouped. The strains were grown for 3 days at 35.5°C on buffered charcoal-yeast extract agar supplemented with L-cysteine. The organisms were harvested, and the inoculum for each strain was prepared in two different buffers [0.1 mol/liter of phosphate buffer (pH 7.0) and 0.05 mol/liter of N-(2-acetoamido)-2-aminoethane sulfonic acid (ACES) buffer supplemented with 4 ml of 10% L-cysteine (pH 7.0) per liter]. The inoculum prepared in each buffer was inoculated into a separate Biolog GN panel and, after incubation, was read manually. The Legionella strains tested yielded biochemical profile variations that allowed distinction between the tested isolates. With the existing data base, the authors found some overlap of biochemical profiles with

Moraxella bovis. The biochemical reactions showed an enhanced color reaction when ACES buffer with L-cysteine was used as diluent for the inoculum. Mauchline and Keevil (73) used the Biolog system to establish a new data base and identify asaccharolytic Legionella spp. They tested single type strains of Legionella pneumophila serogroups 1 through 14 (excluding serogroups 4 and 9) and a single type strain of Legionella bozemanii, Legionella dumoffii, Legionella feeleii, Legionella hacke-

liae, Legionella israelensis, Legionella rubrilucens, Legionella longbeachae, and Legionella micdadei. The isolates were grown on buffered charcoal-yeast extract agar at 37°C for 72 h. Inocula were prepared in Page's amoebal saline, and Biolog plates were inoculated and then incubated at 37°C in either air or a low-oxygen (=4%) atmosphere. The plates were read at 24-h intervals up to 72 h, both visually and with a Merertech Microplate Reader (Atlas Bioscan, Bognor Regis, United Kingdom). When the tested legionellae were incubated in air, some of the reactions were not visible at 24 h and the substrates required longer incubations to turn positive. However, when the same strains were incubated in a reduced-oxygen atmosphere, definite reactions were apparent in 24 h. When tested under these conditions, none of the legionellae had a metabolic profile that closely matched any other bacteria in the Biolog data base, indicating that the profiles obtained were specific for Legionella species. In addition, the authors tested Biolog plates with environmen-

tal isolates that had been provisionally identified by serologic testing as various Legionella species. They compared the results with a combination of the Biolog data base and their own data base and found that the results agreed with those obtained by serologic testing. In further experiments, it was determined by using distilled water as the diluent and a normal aerobic atmosphere during incubation that adequate numbers of positive reactions occurred at 24 h to allow accurate identification of the tested strains. The authors concluded that the Biolog system had the ability to identify the tested Legionella strains at least to the species level but that multiple strains from all known species would have to be characterized to establish a comprehensive and stable data base.

318

STAGER AND DAVIS

Miller and Rhoden (79) evaluated an early version of the Biolog GN data base with a diverse group of clinically relevant members of the family Enterobactenaceae (212 strains) and other gram-negative organisms (105 nonfermenters and 35 oxidase-positive fermenters) consisting of 91 species. The isolates consisted of usual and unusual organisms in proportions likely never to be found in the clinical laboratory and were identified by conventional biochemical and serologic techniques. The Biolog GN was read on the MicroPlate Reader, and only 266 (75.6%) of 352 organisms tested were identified with an acceptable similarity index (SI). Of the 266 strains identified, 87.3% were correct at the genus level and 75.6% were correct at the species level after 24 h. The error rate was 12.8%. In this study of 352 strains, 46.6% of the species were correctly identified at 4 h and 57.1% were correct at 24 h. The error rate was 10.4% at 4 h and 9.6% at 24 h. When the isolates used in this study were reevaluated with an upgrade of the data base (Release 2.00), 86.4% of the members of the family Enterobacteniaceae and 80.4% of the other gram-negative organisms were correctly identified. The authors pointed out that the SIs of the common clinical isolates may be weakened by slow-growing environmental strains. In late 1991, an expanded and revised version of the data base (Release 3.00) was introduced. This version has not yet been evaluated. Recent abstracts have reported on the Biolog GN, but only one of these studies used the MicroPlate Reader. McLaughlin et al. (76) reported on the Biolog GN and MicroPlate Reader for identification of infrequently isolated gram-negative human pathogens. All strains tested were obtained from the American Type Culture Collection. Biolog GN correctly identified 68.5% (89 of 130 strains) to species level and 79.2% (103 of 130 strains) to genus level. Ten percent (13 of 130 strains) were incorrectly identified. Barth et al. (12) evaluated the Biolog GN but did not use the MicroPlate Reader. They tested the Biolog GN with 46 miscellaneous gram-negative bacteria recently isolated from humans. The isolates had been identified by conventional biochemical tests. The Biolog GN results were read at 4 h and at 16 to 24 h. In most cases, the 4-h reading was the same as the later reading. Fifty-nine percent of the organisms were correctly identified to species level, and 83% were identified to genus level. Four of the organisms were listed as the second choice for the Biolog GN, and there was no correlation between conventional and Biolog GN identifications for four other organisms. The colorimetric reactions were difficult to interpret. Roman et al. (100) reported on the Biolog GN but did not use the MicroPlate Reader. They tested 75 strains of P. cepacia isolated from cystic fibrosis patients in Ohio, in Utah, and in Toronto, Canada. The strains had been identified by using conventional biochemicals. At 24 h, 53 (86%) of 62 strains were correctly identified by the Biolog GN. To determine genus similarities and species differences, Wong (124) tested 12 strains of Brucella melitensis, 9 strains of Brucella abortus, and 6 strains of Brucella suis with Biolog plates. Strains that were epidemiologically related (strains isolated from common-source infections and laboratory accidents) were also tested. After incubation, the plates were read visually. Methylpyruvate, monomethyl succinate, and DL-lactic acid were utilized by all Brucella spp. tested. All B. suis isolates tested utilized L-arabinose, a-ketobutyric acid, uronic acid, ,B-hydroxybutyric acid, and uridine. The last two substrates were positive only with B. suis. D-Fructose, alaninamide, L-alanine, L-asparagine, and L-glutamic acid were utilized by all B. abortus isolates tested. It was concluded that each Brucella strain had some distinguishable


features in its metabolic profile and that strains that were epidemiologically related had similar metabolic profiles. However, the results suggested that the Biolog system would not replace standard methods for species identification of Brucella isolates. Ewalt et al. (35) evaluated the manual Biolog system for the ability to differentiate Brucella biovars. The reference system for identification was not mentioned. They tested isolates of B. abortus bv. 19; B. abortus bv. 1, 2, and 4; B. suis bv. 1, 2, and 3; Brucella ovis; Brucella canis; B. melitensis bv. 1 and 2; and Brucella neotomae. There were no visible reactions at 4 h, but there were definite reaction patterns for each Brucella sp. and for the majority of biovars at 24 h. Isolates of the same species and biovar demonstrated some variation in their metabolic profile. Methylpyruvate and DL-lactic acid were utilized by most of the isolates. There is an obvious need for further evaluation of the Biolog GN and MicroPlate Reader for identification of gram-negative bacilli. There have been no published studies of the Biolog GP. With the large Biolog data base for gram-negative and gram-positive bacteria, the Biolog system could enhance the ability of laboratories to identify unusual bacteria. As the capabilities of the Biolog system continue to expand and the accuracies of the new products are verified, the value of the Biolog system in the clinical laboratory should increase. MIDI MICROBIAL IDENTIFICATION SYSTEM Gas chromatography of cellular fatty acids is a rapid and reliable means of identifying organisms encountered in the clinical laboratory (80, 81). Because of the large number of fatty acids found in the cell wall and cell membranes of bacteria and because the composition of cellular fatty acids is a very stable genetic trait that is highly conserved within a taxonomic group, fatty acid composition can be successfully used for identification of bacteria. The MIDI Microbial Identification System (MIS; Microbial ID Inc., Newark, Del.) is a fully automated, computerized, high-resolution gas chromatography system that can analyze more than 300 fatty acid methyl esters ranging in length from 9 to 20 carbons. The MIS computer then searches data bases of known compositions to automatically identify yeasts, anaerobic bacteria, and aerobic bacteria, including mycobacteria. Welch (123) has recently reviewed the applications of cellular fatty acid analysis in clinical microbiology and has described the fatty acid profiles found in various microorganisms. The MIS data base includes information for identification of 15 genera and 65 species or subspecies of the family Enterobacteriaceae; 45 Pseudomonas species, subspecies, or biovars; 18 Staphylococcus species or subspecies; 19 Bacillus species; and 53 additional genera of aerobic bacteria containing 197 species, subspecies, or biovars. In addition, the MIS data base includes information for identification of 32 species, subgroups, or complexes of mycobacteria; 31 genera of anaerobes containing 254 species, types, or groups; and 23 genera of yeasts including 195 species or subspecies. Periodically, an expanded and updated data base is provided at no cost to users of the system. The MIS is composed of a gas chromatograph (model 5890A; Hewlett-Packard Co., Avondale, Pa.) equipped with a fused-silica capillary column (25 m by 0.2 mm [inner diameter]) containing cross-linked methyl-phenyl silicone as the stationary phase, an automatic injector, a sample controller, a sample tray, a flame ionization detector, an elec-

VOL. 5, 1992

tronic integrator controlled by a DOS-based 386 computer such as the Hewlett-Packard Vectra, and a ThinkJet Printer/ Plotter. The MIS software includes programs for operation of the gas chromatograph, automatic peak naming, data storage, and comparison of the unknown profile with one or more data bases by using pattern recognition algorithms. The data base contains more than 100,000 profiles of strains collected worldwide and grown under standardized conditions. Generally, 20 or more strains of a species or subspecies are included in the data base. Environmental aerobic bacteria are cultured on Trypticase soy broth agar (TSBA) at 28°C. Bacteria that grow poorly on this medium are grown on a more enriched medium (e.g., Neissenia spp. are grown on chocolate agar). Clinical aerobic bacteria are incubated at 35°C on TSBA. Mycobacteria are grown at 35°C on Middlebrook 7H10 or 7H11 agar with oleate-albumin-dextrose-catalase enrichment (Mycobacterium marinum is grown at 30°C). Yeasts are cultured on Sabouraud dextrose agar at 28°C. Anaerobic bacteria are grown overnight in peptone-yeast extract-glucose broth at 35°C and harvested by centrifugation. Approximately 40 mg of cells is saponified with 1.0 ml of 1.2 M NaOH in 50% aqueous methanol by heating the cells in a boiling-water bath for 30 min. The saponified cellular lipids are methylated with 2 ml of methylation reagent (325 ml of 6 N HCI and 275 ml of methanol) for 10 min at 80°C, and the fatty acids are extracted with 1.25 ml of extraction reagent (200 ml of hexane and 200 ml of methyl tert-butyl ether) by rotating the mixture for 10 min at room temperature. The lower, aqueous phase is removed; 3 ml of sample cleanup reagent (10.8 g of NaOH dissolved in 900 ml of distilled water) is added; and the tube is rotated for 5 min. About two-thirds of the organic phase is then transferred to a septum-capped sample vial and placed in the sample tray. Samples are then logged into the computer. The automatic injector injects 2,ul of the extract through a heated, self-sealing rubber septum in the heated injection port (250°C). The autosampler allows the system to be operated unattended for up to 2 days at a time. The injected sample is volatilized and swept through the column by a stream of carrier gas (hydrogen). The column is encased in a thermoregulated oven, and the MIS computer raises the temperature from 170 to 270°C at 5°C per min. At the end of the analysis, the column is cleaned by heating (310°C for 2 min). The flame ionization detector (300°C) sends the electronic signals produced by the analytes to integrators that amplify and process the signals. These data are passed to the computer, where they are stored and can be compared with the data base. MIS identifications are listed with a confidence measurement (SI) on a scale of 0 to 1.0. The total test time for each specimen is 30 min. The calibration standard used with the MIS is a mixture of straight-chain saturated fatty acids from 9 to 20 carbons in length and five hydroxyl acids. With the calibration mixture, the retention time of the various peaks can be converted to equivalent chain length data for fatty acid identification. The equivalent-chain-length value for each unknown compound is compared with the external standard for peak naming. Changes in sample injection volume and variables such as carrier gas flow rates and column and detector temperatures will affect the sample retention time. Therefore, the calibration mixture is analyzed after each 10 sample analyses to correct for any possible drift of retention time. The Library Generation Software allows the user to generate data bases. The software contains cluster analysis programs that will generate dendrograms or two-dimensional


319

plots of principal-component analyses. The cluster analysis programs are valuable for tracking nosocomial infections. The MIS has proven valuable in some cases for the differentiation of phenotypically similar organisms and for subgrouping or subspecies characterization of organisms. Wallace et al. (121) used the MIS to compare the fatty acid profiles of Kingella denitrificans, Kingella kingae, and Kingella indologenes with the phenotypically similar Cardiobacterium hominis and Eikenella corrodens. Kingella indologenes, Cardiobacterium hominis, and Eikenella corrodens demonstrated large amounts of cis-vaccenic and palmitic acids, whereas myristic and palmitic acids were the major acids in Kingella denitnificans and Kingella kingae. Only Cardiobacterium hominis lacked 3-hydroxypalmitic acid and 3-hydroxymyristic acid and could be differentiated from Kingella indologenes and Eikenella corrodens by the presence of 3-hydroxypalmitic acid and 3-hydroxymyristic acid, respectively, in these bacteria. Christenson et al. (22) isolated Pseudomonas gladioli from 11 patients with cystic fibrosis and found that it was not associated with any infectious complications. P. gladioli is primarily a plant pathogen and is rarely isolated from humans. Since it resembles P. cepacia, a known infectious agent in cystic fibrosis patients, the authors confirmed the identity of the isolates by using conventional biochemicals, DNA hybridization studies, and analysis of cellular fatty acid profiles with the MIS. Most of the P. gladioli isolates contained 3-OH C10:0 fatty acids, whereas all 58 strains of P. cepacia lacked this fatty acid. Mukwaya and Welch (84) determined the cellular fatty acid profiles for 42 strains of P. cepacia isolated from patients at five cystic fibrosis centers. Hexadecanoic (C16:0) acid, cis-9 hexadecenoic (C16:1 ci,9) acid, and an isomer of octadecenoic (C18:1) acid were present in significant amounts in all strains, and none of the fatty acids had fewer than 14 carbon atoms. Through numerical analysis of the fatty acid data, the authors identified a different subgroup present at each of the cystic fibrosis centers. Lambert and Moss (65) used the MIS to determine the cellular fatty acid profiles of 182 Legionella strains representing 23 species. The 23 species differed in the relative amounts of 14-methylpentadecanoic (i-C16:0), hexadecanoic (C16:1), and 12-methyltetradecanoic (a-C15:0) acids and could be placed into three major fatty acid groups. When the fatty acid profiles of these Legionella strains were linked with their ubiquinone contents, the strains could be distinguished from other gram-negative bacteria. deBoer and Sasser (31) examined the cellular fatty acid compositions of Erwinia carotovora strains and found that Erwinia carotovora subsp. carotovora and Erwinia carotovora subsp. atroseptica had six common fatty acids but that the subspecies could be differentiated because three of these fatty acids had different ratios. Moss et al. (83) found that Moraxella spp. could be differentiated from Acinetobacter spp., Psychrobacter immobilis, Oligella urethralis, and CDC groups EO-2, EO-3, M-5, and M-6 on the basis of differences in cellular fatty acids. The MIS also determined that Moraxella bovis, Moraxella nonliquefaciens, and some strains of Moraxella lacunata have a common fatty acid group, while Moraxella osloensis, Moraxella phenylpyruvica, Moraxella atlantae, and strains of Moraxella lacunata have species-specific fatty acid profiles. They also used pigment production, cellular morphology, and cellular fatty acid profiles of strains originally classified as CDC group EO-2 to identify them as Psychrobacter immobilis, EO-3, or EO-2.

320

STAGER AND DAVIS

Osterhout et al. (91) evaluated the MIS with 573 isolates of gram-negative nonfermentative bacteria including Pseudomonas, Shewanella, Comamonas, Flavimonas, Xanthomonas, Acinetobacter, Agrobacterium, Alcaligenes, Bordetella, Flavobacterium, Methylobacterium, Moraxella, Ochrobactrum, CDC group EO-2, CDC group VB-3, CDC group M-5, CDC group IVC-2, Chryseomonas, Sphingobactenum, Oligella, and Weeksella species. Of these, 536 were fresh clinical isolates and 37 were reference strains. All isolates were identified by conventional tests. Isolates were cultured at 28°C for 22 to 26 h on TSBA with 5% sheep blood. MIS identifications with an SI of 20.5 were considered a good match. The MIS correctly identified 478 (90%) of 532 strains contained in the data base. However, only 314 (59%) had an SI of .0.5. Of the 54 strains incorrectly identified, 26 were Acinetobacter, Moraxella, orAlcaligenes strains and 12 were Pseudomonas pickettii strains. Of 41 isolates (representing 12 species) not in the data base, 33 were not identified or were named with very low SIs. The authors attributed discrepancies to incorrect or poorly defined data base profiles or an inability to differentiate species that are genetically and chemically closely related. Reference strains of P. aeruginosa and X. maltophilia demonstrated a significant variation in SIs when they were tested on 100 different occasions. There was significant improvement in SIs when organisms were incubated at 35°C and analyzed by a data base generated at this temperature with the Library Generation Software. It was concluded that the development of a data base for the culture conditions routinely used in clinical microbiology laboratories might improve the accuracy of the MIS. Abstracts concerning the identification of common and uncommon clinical isolates by the MIS have recently been published. deTurck et al. (33) evaluated the MIS with well-characterized bacterial strains. The MIS correctly identified 102 (93.4%) of 109 strains of Pseudomonas spp. (including 13 species) and 58 (81.2%) of 71 strains of other gram-negative nonfermenters (including 12 species). None of the strains were incorrectly named; rather, they produced no identification. Of 30 strains of Bacteroidesffragilis tested, 29 (97%) were correctly identified. The authors reported that 118 additional anaerobic bacteria of various genera were all correctly identified to the genus level. The MIS, using the Virginia Polytechnic Institute anaerobe library, was compared with the RapID ANA II System (Analytab Products) for identification of anaerobic clinical isolates (74). Discrepancies between the two systems were resolved at Virginia Polytechnic Institute by conventional testing. The isolates tested were 25 Bacteroides spp., 8 other anaerobic gramnegative bacilli, 20 Clostndium spp., 10 nonsporeforming gram-positive bacilli, and 14 anaerobic gram-positive cocci. The MIS and the RapID ANA System identified 97 and 94%, respectively, of the anaerobic isolates tested (P > 0.20). Ayers and Solomon (6) cultured Staphylococcus strains on TSBA at 28°C (595 cultures) or 5% sheep blood agar at 350C (1,318 culture). The standard method for identification of isolates was not mentioned. The strains were chromatographed for cell wall fatty acids in the MIS, and the data were stored by Library Generation Software. Cluster analysis with two-dimensional plots and dendrograms were used to demonstrate isolate relationships. Factors that adversely affected relatedness included growth on TSBA, small-colony variants, the initial culture from frozen or lyophilized reference strains, cell wall fatty acid heterogeneity with some species, and the similar taxonomic position of some species. The MIS software recognized tight clusters with high SIs


when full-bodied colonies were obtained from 5% sheep blood agar that had been incubated at 35°C for 24 h. Euclidian distance cuts were used to define groups, and reference strains or phenotypes were used to name the defined groups. The authors found that all 29 named species were identified by the generated library and 20 unnamed staphylococcal strains (clusters) were recognized. Moss et al. (82) used the MIS to create cellular fatty acid library entries for 358 bacterial species or unnamed groups commonly recovered from clinical specimens. There were 91 cellular fatty acid groups, of which some were specific at the genus or species level while others contained species from different genera. Conventional culture and biochemical methods served as the reference identification for the isolates. Of 2,018 isolates, the MIS correctly identified 78% to the proper cellular fatty acid group, 13% were equivocal because of one or more unusual biochemical reactions, 6% were unidentified by both cellular fatty acid and conventional methods, and 3% were misidentified by cellular fatty acids. Master et al. (71) identified 13 isolates of pinkpigmented, oxidase-positive bacteria with the MIS, API 20E, Rapid NFT, BBL Sceptor (Becton Dickinson), Corning Uni-NF Tek (Flow Laboratories, Inc., Rosalyn, N.Y.), MicroScan, Pasco (Difco, Inc., Wheat Ridge, Co.), and Vitek GNI. Identifications of the same isolates by conventional biochemical methods served as reference identifications. The unknown fatty acid profiles were analyzed by both the routine and an experimental MIS data base. Ten Methylobactenum spp. and one Pseudomonas vesicularis isolate were correctly identified, and one isolate was correctly classified as a gram-positive bacillus by the MIS. The remaining isolate was not identified. The commercial biochemical systems were unable to identify any of the 13 isolates. Extensive evaluations of the MIS for the accurate identification of common and uncommon clinical isolates (aerobes, mycobacteria, yeasts, and anaerobes) and for isolates which are misidentified, unidentified, or identified with a low likelihood by commonly used commercial systems will be required to determine the utility of the MIS in the routine clinical setting.

AUTOSCEPTOR The autoSceptor (Becton Dickinson) will automatically read and report up to 15 Sceptor panels that have been incubated off-line. The autoSceptor is capable of identifying gram-negative bacilli and interpreting breakpoint or MIC tests on gram-positive and gram-negative bacteria after 18 to 24 h of incubation. Various panel formats are available for susceptibility testing. The identification panel contains 24 dried, modified, conventional substrates. The data base includes information for identification of 42 species of the family Enterobactenaceae and 36 groups, genera, or species of nonfermentative and oxidase-positive gram-negative bacilli in 18 to 24 h. The autoSceptor is composed of a modified InterMed ImmunoReader NJ-2000 microELISA reader, Digital Professional 350/380 computer, and a Data Management Center (DMC). The DMC has menus for demographic entry, data analysis, editing, report generation, and epidemiological evaluations. A bidirectional mainframe interface is available. Inocula in 0.85% saline are prepared by suspending colonies grown overnight on selective or nonselective media to the equivalent of a 0.5 McFarland turbidity standard. Panels are inoculated (100 ,ul per well) with a computer-controlled,

VOL. 5, 1992

321


TABLE 5. Studies comparing automated identification systems No. of organisms tested

System (% identification accuracy)

Vitek

autoSCAN-4

WalkAway-96

358a

GNI

(94.7)

R-GNB (94.3) R-GNB (95.2) R-GNB (95.6) R-GNB (86.0) R-GNB (82.0) R-GNB (83.0)

142b

GNI

(79.6)

R-GNB (74.6)

232

YBC (85.0)

246 292a

91b

GNI (99.2) GNI (96.2) GNI (90.1)

180a

60b

a b

YIP (59.0)

Members of the family Enterobacteriaceae. Gram-negative organisms that were not members of the

P value

Reference

0.50 >0.10 >0.05 0.05 (GNI vs AP80) >0.50 (R-GNB vs AP80) k

z z

C)e

'413

0~~~~~~ C)

t' 1 n-C u nn o o o o o t z zZ zo

o

0m

d- d00

d1 :

0 ~

~

o~~~~

~

Cu

C)

0 0 0 CZ

0

*--

.0U

t

ZO Z ZZ

00 en

zzo

o

ZZ

zo >Z

azoZ,

>.ZR

CD

z 0

0U

V°

Q

L

z

E

Cu

Cu)

..

'000

0

(0

C.)

C)

.o

0

4-o 0

C)

c0 C)

r.

HE

0

zo 0o

0

C) (A

u

co2

10 la C)

0

Cu3

Cu 0

C.

'4

00

0 0 0=zzzzzz0

E

Cu ._

o o

"

"D oo 00

Cu

a-'i

t) t

co

ct

0

0

o

rCuCW

*00

;

0

0

0

z

3

003

rA

0 0

0

C._

IC,"-q zt

oy~~~~~~

'IC00000

C.)

o~~~~~~o

o

0-

14

._ C.)

I1

z

0

I4

Cu

.0

z

0

C

3

00

009

-

-

i

-

-

~~~~~~-

0

000 Cu

D

80m

.D d ,It ~

~ ''

Z ooo

0

'

o°

pe

0 IR

0

"It

It

Cu.

.0

C14~~~~~~~~~~C (=10 0

0-4 *-

aC;

00

t

tr,o

zz

Z

4-

0

o

00

d-i

>4 tt

z

t 0

14 u ;t.

I

> >.

>, >

w 00

0

Z

_

m~-

(

U

w

0

>

0

Cu~~~=

4-0 ~ ~

~

~

~

~

~

~

tn

D'4

VOL. 5, 1992

tion for what constitutes an accurate or inaccurate test result for automated identification systems and under what circumstances they should be tested, the overall accuracy of the systems is difficult to determine. Miller (78) presents a very good description of steps which could lead to a consensus approach to resolve many of the issues that have plagued us during our review of the literature on automated identification systems. We are unable to objectively answer the question of which instrument is best for which application. Among the many reasons is that the accuracy of a system is highly dependent on the organisms tested and that these populations are never the same in the various studies. Also, by the time a study is reported, the manufacturer has often changed the data base, altered thresholds on the reader, or modified, added, or replaced some of the substrates. In addition, some studies have used photometrically standardized inocula, whereas others have used generally less reliable visually prepared inocula. Also, the reference system in some studies may not always yield the correct identification, and hence the system under study can be unfairly penalized. In a recent reevaluation of the API 20E, O'Hara et al'. (89) showed that a system did not even necessarily provide the same level of accuracy over time. They were not able to explain why this was so, but an obvious difference was an expanded data base. The list of variables goes on! However, it is obvious from the literature that automated systems accurately identify common clinical isolates. Rare biotypes of common organisms and unusual organisms are often misidentified, identified at low likelihood, or not identified. The accuracy of systems has been observed to vary even with the same kind of organism when it is evaluated in different laboratories. Kelly et al. (60), Stevens et al. (111), and Truant et al. (120) suggested that variation in the biotypes of individual species encountered in different areas of the country may contribute to this performance variation. In support of this hypothesis, Kelly et al. (60) reported a collaborative evaluation of the Autobac by four laboratories in various parts of the country. Each laboratory tested organisms collected in its geographic area. Correct identification of Citrobacter spp. ranged from 69 to 100% among the laboratories, and that of Enterobacter spp. ranged from 62 to 100%. The reason for this variation was not evident. The authors also noted a variation in accuracy with Acinetobacter spp. at two of the laboratories. One of these laboratories correctly identified 65% of their strains, whereas the other correctly identified 100% of their strains. When the strains from the first laboratory were retested at the second laboratory, the accuracy forAcinetobacter spp. was similar to that of the first laboratory. Land et al. (67) reported that only 27% of serogroup D isolates of Cryptococcus neoformans, which are common to Europe and other temperate regions, were correctly identified by the MicroScan YIP. On the other hand, 83% of serogroups B and C, serogroups geographically found nearest to the manufacturer, were correctly identified. Serogroup A isolates, which are commonly found in the remainder of the United States, were identified 73% of the time. The results of Kelly et al. (60) and Land et al. (67) suggest that the data bases of these automated identification systems include limited biotypes from certain geographic areas. That geographic or regional variation exists is not surprising, but apparently some manufacturers have not adequately regionalized the strains in their data base, and this has caused problems. We know that several manufacturers have programs for acquisition of biotypes of common and unusual organisms from various geographic areas of the


323

country. As the data bases are expanded to include these organisms, the accuracy of the systems should improve. Would more substrates in a panel improve the accuracy of identification? Lapage et al. (68) designed a computer program to address this question. The results indicated that aberrant strains of members of the family Enterobacteriaceae and nonfermentative gram-negative bacilli required an average of 29 to 32 tests for reliable identification and that additional tests did not improve the accuracy. Most of the systems reviewed in this paper contain 29 or more substrates in their identification system. Certainly, there are circumstances under which the availability of fewer than 29 substrates is limiting. For example, the Avantage BIC, with 20 substrates, will identify only 10 members of the gramnegative organisms that are not members of the Enterobacteriaceae. How effective are automated identification systems in determining the relatedness of isolates for epidemiologic purposes? The Biolog system and the MIS have cluster analysis programs in the form of dendrograms and twodimensional or three-dimensional plots to demonstrate the relatedness of strains or species. There are, however, only limited reports on the effectiveness of these systems for epidemiologic purposes. Wong (124) used the Biolog system to test Brucella strains which were isolated from commonsource infections or laboratory accidents and found that strains that were epidemiologically related did have similar metabolic profiles. However, it was not indicated whether cluster analysis or dendrograms were used in that study. The study suggested that the Biolog system should prove useful in epidemiologic studies. Mukwaya and Welch (84) used the MIS to determine the cellular fatty acid profiles for 42 strains of P. cepacia isolated from patients at five cystic fibrosis centers and found, through numerical analysis of the data, that a different subgroup was present at each of the centers. Clarridge and Harrison (23) evaluated the MIS for strain differentiation of X. maltophilia from the surgical intensive care unit (7 strains) and medical intensive care unit (9 strains) in a hospital. When it was assumed that a group of strains from different patients were clustered with the same degree of relatedness as defined by all strains from one patient during different extractions and runs on the MIS, the medical intensive care strains made up a single group whereas the surgical intensive care strains were grouped into three clusters. Only two studies have evaluated the reproducibility of biochemical tests with automated instruments. Stoakes et al. (113) retested 50 strains of anaerobes (30 gram-negative bacilli and 20 clostridia) with the autoSCAN-4 AIP. Of the 1,200 reactions, 69 (5.6%) were recorded differently between the two trials. As a result, 8% of the strains changed classification from correct identification to incorrect identification. Murray et al. (85) evaluated the reproducibility of the biochemical reactions obtained with the Quantum II BIC by testing, on three consecutive days, 40 gramnegative organisms belonging to members of the family Enterobacteriaceae. Identical biocodes for all three tests were obtained for only 10 (25%) of the 40 organisms. After modification of the photometer, an additional 25 isolates were evaluated. For the triplicate tests, identical biocodes were obtained for 13 organisms (52%). The results of these last two studies suggest that the automated systems evaluated are not highly reproducible in generating identical biocodes and that the error rate would be too great for the systems to be of value in epidemiological studies. How reliable other automated systems would be in generating identical biocodes has not been reported in the literature.

324

STAGER AND DAVIS

We believe that more standardized evaluations of automated identification systems should be performed but that anyone contemplating such studies should examine very carefully the observations of Miller (78). When trying to compare or determine the accuracy of the available instruments, we were surprised to find that comprehensive evaluations have not been reported, particularly since many of the changes in data bases and other improvements have been made. Some automated systems or parts of systems have not been independently evaluated. Manufacturers should consider providing test cards or panels that contain only the supplemental substrates required for their identification systems. When supplemental tests are necessary,

they are frequently not readily available,

and significant delay and expense are required for identification of what should not be a difficult isolate. There is a need to develop techniques to identify and determine susceptibility patterns for life-threatening infections more quickly. We need methods for direct testing of positive blood cultures, for example. We should make better use of the computer capabilities of the systems. Epidemiology, determination of the significance of organisms identified, and continuing education could easily be enhanced. For example, when a relatively rare isolate is identified, the operator should be able to query the computer data base for information regarding its likelihood of causing infection, its probable susceptibility pattern, and other relevant data. Although it is difficult to imagine a more exciting and stimulating period for clinical microbiologists than the recent past, the immediate future appears to be at least as stimu-

lating.

ACKNOWLEDGMENTS We thank Janice

manuscript.

Edwards-Bryant and Susan Fogg for typing the REFERENCES

1.

Aldridge, K. E., J. H. Wall, R. Facklam, L. B. Reller, A. Janney, and C. V. Sanders. 1988. Identification of viridans streptococci by three commercial systems, abstr. C-346, p. 389. Abstr. 88th Annu. Meet. Am. Soc. Microbiol. 1988. American Society for Microbiology, Washington, D.C. 2. Almeida, R. J., J. H. Jorgenson, and J. E. Johnson. 1983. Evaluation of the AutoMicrobic system gram-positive identification card for species identification of coagulase-negative staphylococci. J. Clin. Microbiol. 18:438-439. 3. Appelbaum, P. C., M. R. Jacobs, J. I. Heald, W. M. Palko, A. Duffett, R. Crist, and P. A. Naugle. 1984. Comparative evaluation of the API 20S system and the AutoMicrobic system gram-positive identification card for species identification of streptococci. J. Clin. Microbiol. 19:164-168. 4. Armon, R., I. Potasman, and M. Green. 1990. Biochemical fingerprints of Legionella spp. by the BIOLOG system: presumptive identification of clinical and environmental isolates. Lett. Appl. Microbiol. 11:290-292. 5. Aure, M., J. Brisker, M. Hirschmann, R. Almazan, and J. Keiser. 1988. Comparative evaluation of the Quantum II and the API 20C systems for identification of germ tube negative yeasts, abstr. C-305, p. 382. Abstr. 88th Annu. Meet. Am. Soc. Microbiol. 1988. American Society for Microbiology, Washington, D.C. 6. Ayers, L. W., and M. C. Solomon. 1991. Performance of the Library Generation Software of the MIS system in the management of Staphylococcus speciation, abstr. C-156, p. 368. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 7. Barry, A. L., and R. E. Badal. 1982. Identification of Enterobacteriaceae by the AutoMicrobic system: Enterobac-


teriaceae biochemical cards versus Enterobacteriaceae-plus biochemical cards. J. Clin. Microbiol. 15:575-581. 8. Barry, A. L., and R. E. Badal. 1982. Reliability of early identifications obtained with Enterobacteriaceae-plus biochemical cards in the AutoMicrobic system. J. Clin. Microbiol. 16:257-265. 9. Barry, A. L., T. L. Gavan, R. E. Badal, and M. J. Telenson. 1982. Sensitivity, specificity, and reproducibility of the Auto Microbic system (with the Enterobacteriaceae-plus biochemical card) for identifying clinical isolates of gram-negative bacilli. J. Clin. Microbiol. 15:582-588. 10. Barry, A. L., T. L. Gavan, R. E. Badal, and M. J. Telenson. 1982. Direct comparison of two mechanized systems for identification of gram-negative bacilli. Am. J. Clin. Pathol. 78:462470. 11. Barry, A. L., T. L. Gavan, P. B. Smith, J. M. Matsen, J. A. Morello, and B. H. Sielaff. 1982. Accuracy and precision of the Autobac system for rapid identification of gram-negative bacilli: a collaborative evaluation. J. Clin. Microbiol. 15:11111119. 12. Barth, S. S., K. B. Williams, S. J. Gibson, and L. B. Elliott. 1991. Rapid identification of gram-negative bacteria by carbon source oxidation, abstr. C-211, p. 377. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 13. Beckmann, E., and P. Connolly. 1990. Flow cytometry: introduction and microbiological applications. Clin. Microbiol. Newsl. 12:105-112. 14. Belcher, K., N. Warren, and H. P. Dalton. 1990. Evaluation of the MicroScan rapid yeast identification panel (YIP) and API 20C yeast system for the identification of yeasts, abstr. C-152, p. 369. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 15. Bochner, B. 1989. "Breathprints" at the microbial level. ASM News 55:536-539. 16. Bochner, B. R. 1989. Sleuthing out bacterial identities. Nature (London) 339:157-158. 17. Burdash, N. M., G. Teti, M. E. West, E. R. Bannister, and J. P.

Manos. 1981. Evaluation of an automated computerized system (AutoMicrobic system) for Enterobacteriaceae identification. J. Clin. Microbiol. 13:331-334. 18. Burdash, N. M., A. L. Welborn, G. Teti, E. R. Bannister, and J. P. Manos. 1985. Identification of gram-negative bacilli using the Autobac IDX. Diagn. Microbiol. Infect. Dis. 3:59-64. 19. Carnahan, A. M., S. W. Joseph, and J. M. Janda. 1989. Species identification of Aeromonas strains based on carbon substrate oxidation profiles. J. Clin. Microbiol. 27:2128-2129. 20. Castleman, K. R. 1979. Images and digital processing, p. 3-13. In K. R. Castleman (ed.), Digital image processing, 1st ed. Prentice-Hall, Inc., Englewood Cliffs, N.J. 21. Chambers, C., J. Henja, B. Bradley, C. Patterson, C. Kennedy, L. Green, and G. Controni. 1990. Comparison of an automated system (AutoSceptor) to the manual Sceptor system, abstr. C-166, p. 371. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 22. Christenson, J. C., D. F. Welch, G. Mukwaya, M. J. Muszynski, R. E. Weaver, and D. J. Brenner. 1989. Recovery of Pseudomonas gladioli from respiratory tract specimens of patients with cystic fibrosis. J. Clin. Microbiol. 27:270-273. 23. Clarridge, J. E., and E. V. Harrison. 1991. The use of fatty acid analysis for the tracking of Xanthomonas maltophilia (Xm) strains in a hospital, abstr. L-36, p. 428. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 24. Colonna, P., D. Nikolai, and D. Bruckner. 1990. Comparison of MicroScan autoSCAN-W/A, Radiometer Sensititre and Vitek systems for rapid identification of gram-negative bacilli, abstr. C-157, p. 370. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 25. Connelly, M. R., and R. C. Jerris. 1987. Identification of yeasts isolated on antimicrobic-containing media using the Abbott Avantage System, abstr. C-348, p. 381. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Micro-

VOL. 5, 1992

biology, Washington, D.C. 26. Cooper, B. H., S. Prowant, B. Alexander, and D. H. Brunson. 1984. Collaborative evaluation of the Abbott yeast identification system. J. Clin. Microbiol. 19:853-856. 27. Costigan, W. J., and G. E. Hollick. 1984. Use of the Autobac IDX system for rapid identification of Enterobacteriaceae and nonfermentative gram-negative bacilli. J. Clin. Microbiol. 19: 301-302. 28. D'Amato, R. F., H. D. Isenberg, G. A. McKinley, E. J. Baron, R. Tepper, and M. Shulman. 1988. Novel application of video image processing to biochemical and antimicrobial susceptibility testing. J. Clin. Microbiol. 26:1492-1495. 29. Davis, J. R., C. E. Stager, R. D. Wende, and S. M. H. Qadri. 1981. Clinical laboratory evaluation of the AutoMicrobic system Enterobacteriaceae biochemical card. J. Clin. Microbiol. 14:370-375. 30. Debates, M., R. Van Enk, and K. Thompson. 1989. Comparison of MicroScan autoSCAN W/A (rapid panels) with Vitek AMS for identification and susceptibility testing, abstr. C-265, p. 437. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 31. DeBoer, S. H., and M. Sasser. 1986. Differentiation of Erwinia carotovora spp. atroseptica on the basis of cellular fatty acid composition. Can. J. Microbiol. 32:796-800. 32. Denys, G., M. Pfaller, L. Priller, M. Yu, and N. Yamane. 1990. Collaborative evaluation of the autoSCAN-W/A rapid gram neg ID system, abstr. C-159, p. 370. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 33. deTurck, J., D. Brooks, M. Pease, and D. Bobey. 1987. Critical evaluation of computerized gas liquid chromatography (GLC) system for bacterial identification, abstr. C-62, p. 333. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Microbiology, Washington, D.C. 34. El-Zaatari, M., L. Pasarell, M. R. McGinnis, J. Buckner, G. A. Land, and I. F. Salkin. 1990. Evaluation of the updated Vitek yeast identification data base. J. Clin. Microbiol. 28:19381941. 35. Ewalt, D. R., J. B. Payeur, and B. R. Bochner. 1990. Differentiation of Brucella biovars using the Biolog system, p. 73. Abstr. Meet. Am. Assoc. Vet. Lab. Diagnosticians 1990. 36. Facklam, R., G. S. Bosley, D. Rhoden, A. R. Franklin, N. Weaver, and R. Schulman. 1985. Comparative evaluation of the API 20S and AutoMicrobic gram-positive identification systems for non-beta-hemolytic streptococci and aerococci. J. Clin. Microbiol. 21:535-541. 37. Facklam, R., and M. D. Collins. 1989. Identification of Enterococcus species isolated from human infections by a conventional test scheme. J. Clin. Microbiol. 27:731-734. 38. Farnham, S., N. Moss, and J. Scott. 1989. The identification of Vibrionaceae using the Vitek Gram-Negative Identification Card (GNI) with expanded data base, abstr. C-263, p. 437. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 39. Ferraro, M. J. B., M. A. Edelblut, and L. J. Kunz. 1981. Accurate automated identification of selected Enterobacteriaceae at four hours. J. Clin. Microbiol. 13:151-157. 40. Forsythe, W. F., J. A. Brandt, and J. E. Lewis. 1990. Comparison study of Vitek YBC system and API 20C for identification of clinical yeast isolates, abstr. C-155, p. 369. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 41. Freeman, J. W., R. W. Rowland, S. B. Overman, and N. L. Goodman. 1981. Laboratory evaluation of the AutoMicrobic system for identification of Enterobacteriaceae. J. Clin. Microbiol. 13:895-898. 42. Gavini, F., M. 0. Husson, D. Izard, A. Berniguad, and B. Quiviger. 1988. Evaluation of AutoSCAN-4 for identification of members of the family Enterobacteriaceae. J. Clin. Microbiol. 26:1586-1588. 43. Godsey, J., R. Kelley, D. Nothaft, J. Bobolis, G. Enscoe, and K. Tomfohrde. 1988. Evaluation of the MicroScan Rapid Pos ID panel, abstr. C-296, p. 381. Abstr. 88th Annu. Meet. Am. Soc.


325

Microbiol. 1988. American Society for Microbiology, Washington, D.C. 44. Goldstein, J., J. J. Guarneri, P. Della-Latta, and J. Scherer.

1982. Use of the AutoMicrobic and Enteric-Tek systems for identification of Enterobacteriaceae. J. Clin. Microbiol. 15:

654-659. 45. Grasmiek, A. E., N. Naito, and D. A. Bruckner. 1983. Clinical comparison of the AutoMicrobic system gram-positive identification card, API Staph-Ident, and conventional methods in the identification of coagulase-negative Staphylococcus spp. J. Clin. Microbiol. 18:1323-1328. 46. Gregson, D., D. Gopaul, P. Hostetler, H. Bialkowska-Hobrzanska, and 0. Hammerberg. 1991. Evaluation of the Automicrobic System's ability to speciate and detect oxacillin-resistance in coagulase-negative staphylococci, abstr. C-233, p. 381. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 47. Hadfield, T. L., P. Lauderdale, B. Fisher, and R. Tapia. 1989. Comparison of the Vitek AMS and MicroScan Walkaway automated microbial identification systems, abstr. C-264, p. 437. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 48. Hare, D., G. Hall, B. Stinson, and M. Szakal. 1989. Comparative evaluation of the Quantum II, API Yeast-IDENT, and Vitek AMS-YBC systems for rapid identification of yeasts not Candida albicans, abstr. C-300, p. 443. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 49. Hasyn, J. J., and H. R. Buckley. 1982. Evaluation of the AutoMicrobic system for identification of yeasts. J. Clin. Microbiol. 16:901-904. 50. Hasyn, J. J., K. R. Cundy, C. C. Dietz, and W. Wong. 1981. Clinical laboratory evaluation of the AutoMicrobic system for rapid identification of Enterobacteriaceae. J. Clin. Microbiol. 13:491497. 51. Hinnebusch, C., D. Nikolai, and D. Bruckner. 1989. Comparison of API Rapid Strep, IDS Rapid STR System, BBL Minitek Differential Identification System, and Vitek GPI with conventional biochemicals for the identification of viridans group streptococci, abstr. C-259, p. 436. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 52. Hussain, Z., L. Stoakes, D. L. Stevens, B. C. Schieven, R. Lannigan, and C. Jones. 1986. Comparison of MicroScan system with the API Staph-Ident system for species identification of coagulase-negative staphylococci. J. Clin. Microbiol.

23:126-128.

53. Isenberg, H. D., T. L. Gavan, P. B. Smith, A. Sonnenwirth, W. Taylor, W. J. Martin, D. Rhoden, and A. Balows. 1980. Collaborative investigation of the AutoMicrobic system Enterobacteriaceae biochemical card. J. Clin. Microbiol. 11:694702. 54. Johnson, J. E., and A. W. Brinkley. 1982. Comparison of the AutoMicrobic system and a conventional tube system for identification of nonfermentative and oxidase-positive gramnegative bacilli. J. Clin. Microbiol. 15:25-27. 55. Jorgensen, J. H., J. W. Dyke, N. G. P. Helgeson, B. H. Cooper, J. S. Redding, S. A. Crawford, M. T. Andruszewski, and S. A. Prowant. 1984. Collaborative evaluation of the Abbott Avantage System for identification of frequently isolated nonfermentative or oxidase-positive gram-negative bacteria. J. Clin. Microbiol. 20:899-904. 56. Jorgensen, J. H., J. E. Johnson, G. A. Alexander, R. Paxson, and G. L. Alderson. 1983. Comparison of automated and rapid manual methods for the same-day identification of Enterobacteriaceae. Am. J. Clin. Pathol. 79:683-687. 57. Kelley, R. W., S. Bascomb, M. Newland, D. Nothaft, R. Fields, and M. V. Lancaster. 1987. A fluorescent system for two hour identification of gram-negative bacteria, abstr. C-57, p. 332. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Microbiology, Washington, D.C. 58. Kelly, M. T., and J. M. Latimer. 1980. Comparison of the AutoMicrobic system with API, Enterotube, Micro-ID, Micro-

326

59.

60.

61. 62.

63.

64. 65. 66. 67.

68. 69. 70.


STAGER AND DAVIS Media systems, and conventional methods for identification of Enterobactenaceae. J. Clin. Microbiol. 12:659-662. Kelly, M. T., and C. Leicester. 1990. Evaluation of the Auto scan-W/A rapid gram-negative identification-susceptibility testing system, abstr. C-167, p. 371. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. Kelly, M. T., J. M. Matsen, J. A. Morello, P. B. Smith, and R. C. Tilton. 1984. Collaborative clinical evaluation of the Autobac IDX system for identification of gram-negative bacilli. J. Clin. Microbiol. 19:529-533. Kiehn, T. E., F. F. Edwards, D. Tom, G. Lieberman, E. M. Bernard, and D. Armstrong. 1985. Evaluation of the Quantum II yeast identification system. J. Clin. Microbiol. 22:216-219. Kloos, W. E., and C. G. George. 1991. Identification of Staphylococcus species and subspecies with the MicroScan Pos ID and Rapid Pos ID panel system. J. Clin. Microbiol. 29:738-744. Kloos, W. E., and D. W. Lambe, Jr. 1991. Staphylococcus, p. 222-237. In A. Balows, W. J. Hausler, Jr., K. Herrmann, H. Isenberg, and H. J. Shadomy (ed.), Manual of clinical microbiology, 5th ed. American Society for Microbiology, Washington, D.C. Kloos, W. E., and K. H. Schleifer. 1975. Simplified scheme for routine identification of human Staphylococcus species. J. Clin. Microbiol. 1:82-88. Lambert, M. A., and C. W. Moss. 1989. Cellular fatty acid compositions and isoprenoid quinone contents of 23 Legionella species. J. Clin. Microbiol. 27:465-473. Land, G., R. Stotler, K. Land, and J. Staneck. 1984. Update and evaluation of the AutoMicrobic yeast identification system. J. Clin. Microbiol. 20:649-652. Land, G. A., I. F. Salkin, M. ElIZaatari, M. R. McGinnis, and G. Hashem. 1991. Evaluation of the Baxter-MicroScan 4-hour enzyme-based yeast identification system. J. Clin. Microbiol. 29:718-722. Lapage, S. P., S. Bascomb, W. R. Willcox, and M. A. Curtis. 1973. Identification of bacteria by computer: general aspects and perspectives. J. Gen. Microbiol. 77:273-290. Larsson, L. 1991. Gas chromatography and mass spectrometry, p. 153-166. In J. H. Jorgensen (ed.), Automation in clinical microbiology, 1st ed. CRC Press, Inc., Boca Raton, Fla. Lyznicki, J., S. Lester, D. Springer, L. Gvazdinskas, and W. Landau. 1991. Comparison of Microscan and Sensititre panels for the rapid identification and susceptibility testing of gramnegative rods, abstr. C-208, p. 376. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C.

71. Master, R. N., R. L. Sautter, W. J. Brown, J. W. Hong, and A. E. Crist, Jr. 1991. Comparison of the Microbial Identification System with conventional and commercially available systems for the identification of pink-pigmented, oxidasepositive bacteria, abstr. C-218, p. 378. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 72. Matthews, K. R., S. P. Oliver, and S. H. King. 1990. Comparison of Vitek gram-positive identification system with API Staph-Trac system for species identification of staphylococci of bovine origin. J. Clin. Microbiol. 28:1649-1651. 73. Mauchline, W. S., and C. W. Keevil. 1991. Development of the BIOLOG substrate utilization system for identification of Legionella spp. Appl. Environ. Microbiol. 57:3345-3349. 74. McAllister, J. M., R. Master, and J. A. Poupard. 1991. Comparison of the Microbial Identification System and the Rapid ANA II system for the identification of anaerobic bacteria, abstr. C-95, p. 358. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 75. McHugh, L. A., M. T. Clifford, B. A. Basille, and J. A. Washington. 1989. Evaluation of the MicroScan Walk-Away rapid identification and susceptibility testing system, abstr. C-266, p. 437. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C.

76. McLaughlin, J. C., W. G. Barron, T. L. Merlin, J. Fierro, and W. C. Thompson. 1991. A comparison of Biolog, Microscan, and Vitek AMS for the identification of infrequently isolated human gram-negative bacterial pathogens, abstr. C-210, p. 377. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 77. McWhirter, R. H., C. Sand, and P. C. Kibsey. 1990. Comparison of the Vitek YBC System and the Microscan Rapid Yeast ID System for the identification of germ tube negative yeast, abstr. F-12, p. 410. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 77a.Merriam-Webster Inc. 1981. Webster's new collegiate dictionary, 150th anniversary ed., p. 76. The G. & C. Merriam Co., Springfield, Mass. 78. Miller, J. M. 1991. Evaluating biochemical identification systems. J. Clin. Microbiol. 29:1559-1561. 79. Miller, J. M., and D. L. Rhoden. 1991. Preliminary evaluation of Biolog, a carbon source utilization method for bacterial identification. J. Clin. Microbiol. 29:1143-1147. 80. Morgan, M. A. 1984. Gas-liquid chromatography in the clinical microbiology laboratory. Lab. Med. 15:544-550. 81. Moss, C. W. 1981. Gas-liquid chromatography as an analytical tool in microbiology. J. Chromatogr. 203:337-347. 82. Moss, C. W., M. I. Daneshrar, and D. G. Hollis. 1991. Automated GLC computer analysis of cellular fatty acids for identification of bacteria, abstr. C-121, p. 362. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 83. Moss, C. W., P. L. Wallace, D. G. Hollis, and R. E. Weaver. 1988. Cultural and chemical characterization of CDC groups

EO-2, M-5, and M-6, Moraxella (Moraxella) species, Oligella urethralis, Acinetobacter species, and Psychrobacter immobilis. J. Clin. Microbiol. 26:484-492. 84. Mukwaya, G. M., and D. F. Welch. 1989. Subgrouping of 85. 86.

87.

88.

89.

90. 91.

92. 93.

Pseudomonas cepacia by cellular fatty acid composition. J. Clin. Microbiol. 27:2640-2646. Murray, P. R., A. Gauthier, and A. Niles. 1984. Evaluation of the Quantum II and Rapid E identification systems. J. Clin. Microbiol. 20:509-514. Navarro, M. C., M. A. Shulman, J. A. Bahrenburg, L. Klints, and D. Schreier. 1987. Comparison of ALADIN video image processing to manual interpretation of An-IDENT, abstr. C-59, p. 333. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Microbiology, Washington, D.C. Nea, L., L. Van Pelt, S. Bascomb, K. Tomfohrde, D. W. Rogers, W. A. Sybers, and K. Irmen. 1989. Evaluation of the Micro Scan rapid ID/MIC system, abstr. C-267, p. 438. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. Oblack, D. L., J. C. Rhodes, and W. J. Martin. 1981. Clinical evaluation of the AutoMicrobic system yeast biochemical card for rapid identification of medically important yeasts. J. Clin. Microbiol. 13:351-355. O'Hara, C. M., D. L. Rhoden, and J. M. Miller. 1992. Reevaluation of the API 20E identification system versus conventional biochemicals for identification of members of the family Enterobacteriaceae: a new look at an old product. J. Clin. Microbiol. 30:123-125. O'Hara, C. M., D. L. Rhoden, and P. B. Smith. 1990. Agreement between visual and automated UniScept API readings. J. Clin. Microbiol. 28:452-454. Osterhout, G. J., V. H. Shull, and J. D. Dick. 1991. Identification of clinical isolates of gram-negative nonfermentative bacteria by an automated cellular fatty acid identification system. J. Clin. Microbiol. 29:1822-1830. Peterson, E. M., J. T. Shigei, A. Woolard, and L. M. de la Maza. 1988. Identification of viridans streptococci by three commercial systems. Am. J. Clin. Pathol. 90:87-91. Pfaller, M. A., M. J. Bale, K. R. Schulte, and F. P. Koontz. 1986. Comparison of the Quantum II bacterial identification system and the AutoMicrobic system for the identification of gram-negative bacilli. J. Clin. Microbiol. 23:1-5.

VOL. 5, 1992

94. Pfaller, M. A., T. Preston, M. Bale, F. P. Koontz, and B. A. Body. 1988. Comparison of the Quantum II, API Yeast Ident, and AutoMicrobic systems for identification of clinical yeast isolates. J. Clin. Microbiol. 26:2054-2058. 95. Pfaller, M. A., D. Sahm, C. O'Hara, C. Ciaglia, M. Yu, N. Yamane, G. Scharnweber, and D. Rhoden. 1991. Comparison of the AutoSCAN-W/A rapid bacterial identification system and the Vitek AutoMicrobic system for identification of gramnegative bacilli. J. Clin. Microbiol. 29:1422-1428. 96. Plorde, J. J., J. A. Gates, L. G. Carlson, and F. C. Tenover. 1986. Critical evaluation of the AutoMicrobic system gramnegative identification of glucose-nonfermenting gram-negative rods. J. Clin. Microbiol. 23:251-257. 97. Quinn, P. R., and C. D. Horstmeier. 1990. Evaluation of Vitek Yeast Biochemical Card (YBC) with an expanded data base for the identification of clinically important yeasts and yeastlike organisms, abstr. F-11, p. 410. Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 98. Rhoden, D. L., and C. M. O'Hara. 1989. Evaluation of the updated QUANTUM II system for the identification of gramnegative bacilli. J. Clin. Microbiol. 27:2420-2422. 99. Rhoden, D. L., P. B. Smith, C. N. Baker, B. Schable, and S. Stocker. 1985. AutoSCAN-4 system for identification of gramnegative bacilli. J. Clin. Microbiol. 22:915-918. 100. Roman, S. B., L. A. Carson, C. M. O'Hara, D. A. Pegues, and M. Miller. 1991. Comparison of four identification methods used to identify Pseudomonas cepacia isolated from sputum of cystic fibrosis patients, abstr. C-222, p. 379. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 101. Ruoff, K. L., M. J. Ferraro, M. E. Jerz, and J. Kissling. 1982. Automated identification of gram-positive bacteria. J. Clin. Microbiol. 16:1091-1095. 102. Salkin, I. F., K. H. Schadow, L. A. Bankaitis, M. R. McGinnis, and M. E. Kemna. 1985. Evaluation of Abbott Quantum II yeast identification system. J. Clin. Microbiol. 22:442-444. 103. Schiminsky, N., and P. Ferrieri. 1987. The use of the Vitek AMS for speciation of the viridans and group D streptococci, abstr. C-94, p. 339. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Microbiology, Washington, D.C. 104. Sekhon, A. S., A. A. Padhye, A. K. Garg, and W. R. Pruitt. 1987. Evaluation of the Abbott Quantum II yeast identification system. Mykosen 30:408-411. 105. Shulman, M. A., M. C. Navarro, J. A. Bahrenburg, S. Jahnke, and D. Schreier. 1987. Comparison of the ALADIN video image processing to manual interpretation of UniScept 20E, abstr. C-58, p. 333. Abstr. 87th Annu. Meet. Am. Soc. Microbiol. 1987. American Society for Microbiology, Washington, D.C. 106. Sielaff, B. H., J. M. Matsen, and J. E. McKie. 1982. Novel approach to bacterial identification that uses the Autobac system. J. Clin. Microbiol. 15:1103-1110. 107. Simmonds, J. E., M. R. Motyl, and J. C. McKitrick. 1990. Evaluation of Vitek Yeast Biochemical Card (YBC) with expanded data base and two rapid enzymatic panels: Micro scan Yeast Identification Panel (YIP) and API Yeast-Ident (Y-I) systems for identification of medically important yeasts, abstr. 1068, p. 264. Program Abstr. 30th Intersci. Conf. Antimicrob. Agents Chemother. American Society for Microbiology, Washington, D.C. 108. Smith, S. M., K. R. Cundy, G. L. Gilardi, and W. Wong. 1982. Evaluation of the AutoMicrobic system for identification of glucose-nonfermenting gram-negative rods. J. Clin. Microbiol. 15:302-307. 109. Snyder, J., G. Buck, and B. Nohinek. 1990. Evaluation of the Abbott Avantage for bacterial identification, susceptibility testing, and impact on antibiotic selection, abstr. C-172, p. 372.


327

Abstr. 90th Annu. Meet. Am. Soc. Microbiol. 1990. American Society for Microbiology, Washington, D.C. 110. Staneck, J. L. Personal communication. 111. Stevens, M., R. K. A. Feltham, F. Schneider, C. Grasmick, F. Schaak, and P. Roos. 1984. A collaborative evaluation of a rapid automated bacterial identification system: the Autobac IDX. Eur. J. Clin. Microbiol. 3:419-423. 112. St.-Germain, G., and D. Beauchesne. 1991. Evaluation of the MicroScan rapid yeast identification panel. J. Clin. Microbiol. 29:2296-2299. 113. Stoakes, L., T. Kelly, K. Manarin, B. Schieven, R. Lannigan, D. Groves, and Z. Hussain. 1990. Accuracy and reproducibility of the MicroScan rapid anaerobe identification system with an automated reader. J. Clin. Microbiol. 28:1135-1138. 114. Stoakes, L., B. C. Schieven, E. Ofori, P. Ewan, R. Lannigan, and Z. Hussain. 1992. Evaluation of MicroScan Rapid Pos Combo panels for identification of staphylococci. J. Clin. Microbiol. 30:93-95. 115. Stockman, L., and G. D. Roberts. 1991. Evaluation of MicroScan Rapid Yeast Identification Panel (version 17), abstr. F-23, p. 412. Abstr. 91st Gen. Meet. Am. Soc. Microbiol. 1991. American Society for Microbiology, Washington, D.C. 116. Sylvester, M. K., and J. A. Washington II. 1984. Evaluation of the Quantum II microbiology system for bacterial identification. J. Clin. Microbiol. 20:1196-1197. 117. Tenover, F. C., T. S. Mizuki, and L. G. Carlson. 1990. Evaluation of autoSCAN-W/A automated system for the identification of non-glucose-fermenting gram-negative bacilli. J. Clin. Microbiol. 28:1628-1634. 118. Tomfohrde, K., R. Kelley, D. Nothaft, J. Bobolis, G. Enscoe, and J. Godsey. 1988. Evaluation of the MicroScan Rapid Neg ID Panel, abstr. C-283, p. 379. Abstr. 88th Annu. Meet. Am. Soc. Microbiol. 1988. American Society for Microbiology, Washington, D.C. 119. Tritz, D. M., P. C. Iwen, and G. L. Woods. 1990. Evaluation of MicroScan for identification of Enterococcus species. J. Clin. Microbiol. 28:1477-1478. 120. Truant, A. L., E. Starr, C. A. Nevel, M. Tsolakis, and E. F. Fiss. 1989. Comparison of AMS-Vitek, MicroScan, and Autobac Series II for the identification of gram-negative bacilli. Diagn. Microbiol. Infect. Dis. 12:211-215. 121. Wallace, P. L., D. G. Hollis, R. E. Weaver, and C. W. Moss. 1988. Cellular fatty acid composition of Kingella species, Cardiobacterium hominis, and Eikenella corrodens. J. Clin. Microbiol. 26:1592-1594. 122. Weckbach, L. S., J. L. Staneck, R. C. Tilton, D. Douglas, R. Zabransky, L. Bayola-Mueller, C. M. O'Hara, and D. L. Rhoden. 1990. Abstr. 90th Annu. Meet Am. Soc. Microbiol. 1990, C-234, p. 383. 123. Welch, D. F. 1991. Applications of cellular fatty acid analysis. Clin. Microbiol. Rev. 4:422-438. 124. Wong, J. 1989. Carbon utilization patterns of Brucella sp., abstr. C-210, p. 428. Abstr. 89th Annu. Meet. Am. Soc. Microbiol. 1989. American Society for Microbiology, Washington, D.C. 125. Woolfrey, B. F., R. T. Lally, M. N. Ederer, and C. 0. Quall. 1984. Evaluation of the AutoMicrobic system for identification and susceptibility testing of gram-negative bacilli. J. Clin. Microbiol. 20:1053-1059. 126. Woolfrey, B. F., R. T. Lally, and C. 0. Quall. 1983. Evaluation of the AutoSCAN-3 and Sceptor systems for Enterobacteriaceae identification. J. Clin. Microbiol. 17:807-813. 127. Yamane,

N., F. Koontz, M. Lindsey, and V. Hirst. 1990.

Card (YBC) with Comparison of the Vitek YeastAPIBiochemical 20C for the identification of the updated data base to the Soc. yeast, abstr. C-233, p. 382. Abstr. 90th Annu. Meet. Am.WashMicrobiol. 1990. American Society for Microbiology, ington, D.C.

Risks of "blind" automated identification systems in medical microbiology.

Evaluation of a Fully Automated Research Prototype for the Immediate Identification of Microorganisms from Positive Blood Cultures under Clinical Conditions.

Systems for automated multiple pesticide residue analysis.

Automated donor-recipient identification systems as a means of reducing human error in blood transfusion.

Performance evaluation of three automated identification systems in detecting carbapenem-resistant Enterobacteriaceae.

Author Correction for Hyman et al., Evaluation of a Fully Automated Research Prototype for the Immediate Identification of Microorganisms from Positive Blood Cultures under Clinical Conditions.

UTHealth shared task Track 1.

High throughput automated colorimetric method for the screening of l-lactic acid producing microorganisms.

Automated design of complex dynamic systems.

Automated Scalable Heat Shock Modification for Standard Aquatic Housing Systems.

Multi-scale curvature for automated identification of glaciated mountain landscapes.

identification system.

Privacy, confidentiality and automated health information systems.

PROTAX-Sound: A probabilistic framework for automated animal sound identification.

Automated Identification of Abnormal Adult EEGs.

Identification of beer spoilage microorganisms using the MALDI Biotyper platform.

SLIMM: species level identification of microorganisms from metagenomes.

Proteotyping: Proteomic characterization, classification and identification of microorganisms--A prospectus.

Identification of fungal microorganisms by MALDI-TOF mass spectrometry.

Automated microscopic image analysis for leukocytes identification: a survey.

A PC-based imaging system for automated platelet identification.

New opportunities for the integration of microorganisms into biological pest control systems in greenhouse crops.

Performances and Reliability of Bruker Microflex LT and VITEK MS MALDI-TOF Mass Spectrometry Systems for the Identification of Clinical Microorganisms.

or difficult-to-identify microorganisms isolated from clinical specimens.