Commercial Identification Systems Often Fail To Identify

JOURNAL OF CLINICAL MICROBIOLOGY, Feb. 1988, p. 323-327

Vol. 26, No. 2

0095-1137/88/020323-05$02.00/0 Copyright © 1988, American Society for Microbiology

Commercial Identification Systems Often Fail To Identify Providencia stuartii GIUSEPPE CORNAGLIA,1* BENEDETTO DAINELLI,2 FRANCESCA BERLUTTI,3 AND MARIA CRISTINA THALLER3 Istituto di Microbiologia, Università degli Studi di Verona, Strada Le Grazie, 8, 37134 Verona,' and Dipartimento di Medicina Sperimentale e Scienze Biochimiche, II Università degli Studi di Roma,2 and Istituto di Microbiologia, Università "La Sapienza,"3 Rome, Italy Received 15 July 1987/Accepted 14 October 1987 We tested 145 clinical isolates in an attempt to evaluate some of the most widely used commercial identification systems in Europe in terms of their ability to identify Providencia strains. Two manual miniaturized systems (API 20E and Enterotube II) and three mechanized-automated systems (Cobas-Bact, Sceptor System, and Titertek-Enterobac-Rapid Automated System) were evaluated. Providencia alcalifaciens and Providencia rettgeri strains were correctly identified by all systems in ail cases, and in most cases identification was achieved without the aid of supplementary tube tests. By contrast, Providencia stuartii was identified without the aid of supplementary tube tests for only 42.5% (API 20E), 37.5% (Enterotube), 68.7% (Sceptor), and 71.2% (Cobas-Bact) of the isolates. The overall misidentification rates were 16.3, 11.3, 11.3, and 10%, respectively. The Titertek-Enterobac-Rapid Automated System failed to identify only 1 of 80 strains (1.3%) and required supplementary tests in 2 other cases (2.5%). Since four of the multitest systems examined often failed to correctly identify P. stuartil, we conclude that supplementary conventional tube tests should always be used to distinguish this species from the other taxa of the Proteeae tribe.

MATERIALS AND METHODS

Several reports over the past few years have stressed the increasing importance of the genus Providencia for both clinical microbiologists and clinicians (4, 11, 20). This fact is probably related, to a great extent, to the redefined taxonomy of the Proteeae tribe (4, 15), allowing a more reliable attribution of isolates to definite groups whose biochemical distinction into genera and species has been confirmed by DNA hybridization studies (1, 7-9). A distinctive characteristic of Providencia stuartii strains is that they are often endowed with multiple resistance to antimicrobial drugs, including some of those most recently licensed (2, 4, 5, 15). In the meantime, clinicians have realized that these microorganisms might be responsible for infections more often than was previously supposed (20) and that these infections are not confined to the urinary tract and can be complicated by fatal bacteremias (10, 13, 17). Besides increased resistance to many antimicrobial agents, two further features confirm the importance of P. stuartii strains as emerging nosocomial pathogens: their resistance to disinfectants (4) and their possible major role in the spread of plasmids responsible for resistance to antimicrobial agents (18). All these elements appear to justify fears that infections caused by P. stuartii may constitute an important problem in clinical practice in the near future owing to their increased frequency of isolation and their resistance to antimicrobial agents. In this study, we investigated the accuracy of some commercial identification systems, selected from among those most widely used in European clinical microbiology laboratories, in identifying Providencia species isolated from clinical specimens, with special reference to P. stuartii strains. *

Bacterial strains and reference identification. In our study we used 145 Providencia strains (1 per patient) freshly isolated from a variety of specimens from patients. All the isolates were identified to species level by the following test battery: indole and H2S production; Voges-Proskauer test; urea, gelatin, and esculin hydrolysis; Simmons citrate; phenylalanine deaminase; lysine and ornithine decarboxylase; arginine dihydrolase; fermentation of adonitol, L-arabinose, D-arabitol, cellobiose, dulcitol, erythritol, glycerol, myoinositol, lactose, maltose, D-mannitol, D-mannose, melibiose, a-methyl-D-glucoside, raffinose, L-rhamnose, salicin, sorbitol, sucrose, trehalose, and D-xylose; and the o-nitrotests were phenyl-g-D-galactopyranoside (ONPG) test. Allmethods (6, performed by using test tubes and conventional 16). To calculate the percent identification (12), we developed a matrix of probabilities which included the members of the family Enterobacteriaceae cited by Farmer et al. (3) as taxa and the positivity percentages reported by the same authors for the tests we had performed as the data base. Identifications to species level were also carried out with the reduced battery of tests proposed by Farmer et al. for differentiation within the genus Providencia (3), including urea hydrolysis and fermentations of myo-inositol, adonitol, D-arabitol, trehalose, and D-galactose. The last test could not be included in the matrix we described above, because data on the probabilities of its results were not available for all the species of Enterobacteriaceae. Commercial identification systems. The same strains were also identified by two manual miniaturized systems (API 20E [API Systems S.A., La Balme-les-Grottes, Montalieu-Vercieu, France] and Enterotube Il [Roche Diagnostics, Basel, Switzerland]) and three mechanized-automated systems

Corresponding author. 323

324

J. CLIN. MICROBIOL.

CORNAGLIA ET AL.

tii were less satisfactory than those for P. alcalifaciens and P. rettgeri, especially if we consider the results obtained without the aid of tube tests. The identifications given by API 20E that required no additional tests proved correct for 34 of 80 strains (42.5%) and incorrect for 11 strains; when we performed the supplementary tests required to obtain a higher-confidence identification for a further 35 isolates, 33 strains (41.2% of the total) were correctly identified and 2 were not, thus bringing the misidentification rate of this system up to 16.3%. The identifications given by Enterotube Il proved correct for 37.5% of strains and incorrect for 3.8% without the supplementary tests, which, in turn, yielded a correct identification of 41 strains (51.2%) and an incorrect identification of 6 strains (7.5%). The overall misidentification rate for Enterotube II was thus 11.3%. The mechanizedautomated systems (Sceptor, TTE-RAS, and Cobas-Bact) correctly identified 55 (68.7%), 77 (96.2%), and 57 (71.2%) isolates, respectively, as P. stuartii and indicated P. stuartii as one of the two or three reference organisms most similar to the test isolate in a further 16 (20.0%), 2 (2.5%), and 15 (18.8%) lower-confidence identifications, respectively. Since in most cases these systems leave the microbiologist the choice of performing the additional tests suitable for obtaining higher-confidence identifications, we must assume that all the above identifications would have proved correct, inasmuch as the tests with the highest separation values for the species concerned were the same ones we had used to set up the reference identification system. However, no identification proved incorrect even when the additional tests were suggested by the manufacturer. Thus, the overall misidentification rates for Sceptor, TTE-RAS, and Cobas-Bact were 11.3, 1.3, and 10.0%, respectively. Table 2 lists the P. stuartii strains which were misidentified, along with the wrong identifications. Most of the misidentifications were confined to the Providencia genus or, at most, to the Proteeae tribe. To assess whether these errors might depend on the different percent positivity attributed to the individual tests by different data bases, we checked the identifications on a data matrix developed for each system by excluding from our reference matrix the tests not present in the system considered. In practice, these control matrices yielded the same misidentification rates as the respective systems (data not shown). A few strains were misidentified as species outside the Proteeae tribe, namely Tatumella ptyseos, Yersinia pseudotubercolosis, Yersinia enterocolitica, and Enterobacter agglomerans. Only one strain (PV 48) was misidentified by all the systems: it was an atypical indole-negative, sucrosepositive strain; the reference matrix identified it as P. stuartii or T. ptyseos (with P = 87.6 and 12.7%, respectively); the ultimate diagnosis of this strain as P. stuartii resulted from

(Cobas-Bact [Roche Diagnostics, Basel, Switzerland], ScepCockeysville, Md.], and Titertek-Enterobac-Rapid Automated System [TTE-RAS; Flow Laboratories GmbH, Meckenheim, Federal Republic of Germany]). Cobas-Bact was equipped with the new version (SW8524) of the software program. Inoculations and incubations were performed as specified by the manufacturers. For the manual systems, the test results were converted into profile numbers by the octal coding system and identifications were made through the manufacturer profile indices, which also gave the normalized percent likelihood (P) (12). For three Providencia rettgeri and three P. stuartii isolates, the profile numbers obtained by API 20E were not listed in the API Systems Analytical Profile Index; therefore, they were identified by reference to the computer data base. The mechanized-automated systems gave a printout of their computerized readings and interpretations of the test results, along with P values. Only identifications agreeing with the reference identification at the species level were regarded as correct, and only those disagreeing with the reference identification at that level, even after supplementary tube tests, were considered to be misidentifications. tor System [BBL Microbiology Systems,

RESULTS On the basis of the reference identification, the organisms included in the study could be assigned to all the Providencia species which are of interest in human pathology, namely P. rettgeri (47 strains), P. stuartii (80 strains), Providencia alcalifaciens (16 strains), and Providencia rustigianùi (2 strains). In all cases, the identifications at species level coincided perfectly with those obtained by using the reduced test battery proposed by Farmer et al. (3). The strains identified by the reference matrix as P. alcalifaciens, and only these strains, were all galactose positive. The species P. rustigianii is not included among the taxa of any of the systems evaluated, and as a consequence, a correct identification for the two strains we examined was not achievable. These strains were identified as P. stuartii by Enterotube II and as P. alcalifaciens by all the other systems. Table 1 shows the percentages of correct and incorrect identifications given for the remaining Providencia species by the commercial systems considered. The identifications of P. alcalifaciens and P. rettgeri strains by all the systems generally proved correct and, in most cases, were achieved without the aid of supplementary tube tests. These tests were required only with Enterotube Il for four strains of P. alcalifaciens (25.0% of the total) and with both Cobas-Bact and API 20E for four strains of P. rettgeri (8.5%). The percent correct identifications registered for P. stuar-

TABLE 1. Identification of Providencia strains by the five commercial systems Identification Idsteniiai

API 20E Enterotube Il

No. (%

correct (- STT)r 16 (100.0) 12 (75.0) 16 (100.0) 16 (100.0) 16 (100.0)

P. alcalifaciens No. (%)

correct (+ STT)b

0 (0.0) 4 (25.0) 0 (0.0) 0 (0.0) 0 (0.0)

P. rettgeri

(% No.

No. incorrect 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0)

P. stuartii

c correct (- STT)

correct

No.

No. (%)No(%N.(%N.() No.

(+ STT)

incorrect

(- SIT)

(+ S1T)

( incorrect

43 (91.5) 47 (100.0) 47 (100.0) 47 (100.0) 43 (91.5)

4 (8.5) 0 (0.0) 0 (0.0) 0 (0.0) 4 (8.5)

0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0)

34 (42.5) 30 (37.5) 55 (68.7) 77 (96.2) 57 (71.2)

33 (41.2) 41 (51.2) 16 (20.0) 2 (2.5) 15 (18.8)

13 (16.3) 9 (11.3) 9 (11.3) 1 (1.3) 8 (10.0)

No. (%)

(%

No.

ccorrect

Sceptor TTE-RAS Cobas-Bact a Correct (-STT), Identification agreeing with the reference identification without the aid of supplementary tube tests. b Correct (+STT), Identification agreeing with the reference identification after supplementary tube tests. C Incorrect, Identifications not agreeing with the reference identification even after supplementary tube tests (when required).

correct

IDENTIFICATION OF P. STUARTII

VOL. 26, 1988

325

TABLE 2. Misidentified P. stuartii strains, with specifications of wrong identifications given by the various systems Identification by:

Strain API 20E

Enterotube Il

Sceptor

TTE-RAS

Cobas-Bact

P. vulgaris, M. morganii, P. rettgeri

PV 23

M. morganii

PV 29 PV 34 PV 39 PV 46 PV 47 PV 48

P. alcalifaciens P. rettgeri P. penneri, T. ptyseos P. rettgeri

P. vulgaris

P. penneri

E. agglomerans P. vulgaris

PV 49 PV 51

* *

M. morganii

PV 53 PV 54 PV 70 PV 81

P. alcalifaciens P. alcalifaciens *** M. morganii

PV 84 PV 87

P. alcalifaciens ***

P. alcalifaciens

Y. pseudotubercolosis, P. mirabilis, P. vulgaris

P. vulgaris

Y. enterocolitica

P. rettgeri, M. morganii

E. agglomerans M. morganii

P. vulgaris, M. morganii, P. rettgeri P. alcalifaciens P. alcalifaciens P. rettgeri, M. morganii M. morganii

P. vulgaris, M. morganii, P. rettgeri

*

P. vulgaris, M. morganii, P. rettgeri E. agglomerans

PV 90 PV 91

P. vulgaris, M. morganii, P. rettgeri E. agglomerans

PV 96 PV 98 PV 107 PV 113 PV 140

P. alcalifaciens P. alcalifaciens P. penneri, P. rettgeri ***

PV 146

*

P. alcalifaciens P. alcalifaciens P. vulgaris

P. vulgaris, M. morganii, P. rettgeri P. vulgaris, M. morganii, P. rettgeri

* *

a Asterisks denote a correct identification.

the fact that it was weakly citrate positive after 48 h and glycerol positive and actively motile at 37°C. The last two tests ruled out the diagnosis of T. ptyseos for another strain (PV 113, which was citrate negative within 72 h), which could not be identified by API 20E and Enterotube IL. Finally, a few isolates were misleading because they were D-xylose positive, and these were attributed by Enterotube II to the Enterobacter genus; they were correctly identified by the other systems. DISCUSSION Our results show that P. alcalifaciens and P. rettgeri strains were adequately identified by all the systems examined, even without the aid of supplementary tube tests. The tests included in the standard system configurations alone were sufficient for a correct identification of all P. alcalifaciens strains by API 20E and Cobas-Bact and of all P. rettgeri strains by Enterotube Il. The identifications obtained by Sceptor and TTE-RAS alone proved correct for all the strains of both species. Finally, all the systems assured 100% correct identifications when supplementary tube tests were performed. P. rustigianii is not included among the taxa of any of the systems examined, but it is worth mentioning that the identifications obtained for our two P. rustigianùi isolates accorded with their former taxonomy. Before the formal name P. rustigianii (9) was proposed, these strains were classified as P. alcalifaciens biogroup 3 and were not differentiable by simple tests from P. stuartii strains previously

known as P. alcalifaciens biogroup 4, although they were already considered to belong to a different species (1). Thus, the only true misidentifications were of P. stuartii. Both the rate and nature of the errors varied from system to system. The misidentification rates decrease in the following order: API 20E (16.3%), Sceptor and Enterotube I (11.3%), Cobas-Bact (10.0%), and TTE-RAS (1.3%). As regards the nature of the errors, identifications outside the Proteeae tribe were confined to a few strains that had uncommon biochemical traits and were not easy to identify even by conventional methods. Misidentifications deriving from the inability to differentiate among the three genera within the Proteeae tribe or the three species within the genus Providencia were much more frequent. In some cases, errors were caused by negativity or delayed positivity of tests such as citrate and myo-inositol and could be attributed to a poorer sensitivity of the micromethods (19): most of the strains that were misidentified proved positive by standard tube procedures. These errors enhanced the probabilities that these strains would be identified as Proteus vulgaris, Proteus penneri (8), or ornithine-negative Morganella morganii (7) or, within the genus Providencia, as P. alcalifaciens. When the results of the micromethods agreed with those of the tube tests, the errors could be attributed to the presence or absence in the miniaturized system of the various tests allowing identification of the P. stuartii isolates to both genus and species level. This correlation was investigated by developing a control matrix for each system.

326

J. CLIN. MICROBIOL.

CORNAGLIA ET AL.

Since the frequency of the errors in each matrix was substantially the same as in the respective system (the only difference between the reference and control matrix being the selection of tests), it could be argued that the misidentifications depended on test selection and not on the different percent positivity attributed to the individual tests by the system data bases. Besides urease production, which is tested in all systems and therefore cannot account for substantial differences in resuIts, a correct identification of P. stuartii by the miniaturized systems considered in our study is based mostly on fermentations of polyhydric alcohols (14). Of these carbohydrates, myo-inositol, adonitol, and trehalose allow a differential diagnosis between P. alcalifaciens and P. stuartii. API 20E includes ônly one of them, i.e., myo-inositol, whose results proved the least reliable (see above). This could explain why this system erroneously identified the highest number of P. stuartii strains as P. alcalifaciens, but, since myo-inositol fermentation is of little value in distinguishing P. stuartii from P. rettgeri, the failure to distinguish these two species must also have other causes. The carbohydrates that allowed differentiation between the two species and were included in the systems we considered are D-mannitol, adonitol, rhamnose, and trehalose. P. stuartii strains which are positive in the D-mannitol and adonitol tests can easily be misidentified as P. rettgeri, since the latter species shows a much higher percent positivity. Even a negative rhamnose fermentation test, which is almost mandatory for P. stuartii, cannot contribute very mùch to the ultimate percent identification, because the same reaction is shared by many P. rettgeri strains. Under these conditions and in the absence of a positive control, i.e., a test in which the percent positivity of P. stuartii is greatly superior to that of P. rettgeri, the matrix of probabilities could hardly offer a correct identification. By contrast, the only system including the trehalose fermentation test (TTE-RAS), which is positive for P. stuartii with a very good separation value, showed by far the best results, with a misidentification rate of only 1.3%. In all the cases we have discussed so far, failure to identify P. stuartii was complete; i.e., this species was not included even among lower-confidence identifications. However, correct identification could not be obtained without supplementary tube tests in a much larger number of cases, ranging from 18.7% of overall Cobas-Bact identifications to 51.2% of those by Enterotube II. TTE-RAS again proved the most accurate system, requiring supplementary tests in two cases only (2.5%). In all these cases, the possible identification as P. stuartii often has a low probability, for substantially the same reasons as discussed above for cases of complete misidentification. Thus, this possibility could easily be overlooked by both laboratory and ward staff, mainly because awareness of the possible role of P. stuartii as a nosocomial pathogen is probably not yet sufficiently strong to prompt the use of identification techniques that are, as in the case of the individual tube tests, in sharp contrast to the present tendency to prefer rapid unitary procedures. However, we think that at least the most frequent errors of the multitest systems should be eliminated by checking the isolates identified as Proteus penneri and ornithine-negative M. morganii and by confirming all diagnoses of mannitolnegative P. rettgeri and adonitol-negative P. alcalifaciens. Owing to the frequency of incorrect or low-confidence findings, it would be even better to carry out tube tests that have good separation values between species of the tribe Proteeae whenever an isolate is thought to belong to this tribe. These tests, selected in accordance with those advo-

cated by Farmer et al. (3), should be performed simultaneously with the tests already present in the miniaturized systems to avoid misidentification and a 24-h delay in obtaining high-confidence responses. ACKNOWLEDGMENTS We thank all those who sent us Providencia isolates, Giuseppe Satta for his helpful comments, and Anthony Steele for his help with the English version of this paper.

1.

2.

3.

4. 5. 6.

7.

8.

9.

10. 11.

12. 13. 14.

15.

16.

LITERATURE CITED Brenner, D. J., J. J. Farmer III, G. R. Fanning, A. G. Steigerwalt, P. Klykken, H. G. Wathen, F. W. Hickman, and W. H. Ewing. 1978. Deoxyribonucleic acid relatedness of Proteus and Providencia species. Int. J. Syst. Bacteriol. 28:269-282. Cornaglia, G., R. Pompei, B. Dainelli, and G. Satta. 1987. In vitro activity of ciprofloxacin against aerobic bacteria isolated in a Southern European hospital. Antimicrob. Agents Chemother. 31:1651-1655. Farmer, J. J., III, B. R. Davies, F. W. Hickmann-Brenner, A. C. McWorther, G. P. Huntley-Carter, M. A. Asbury, C. Riddle, H. G. Wathen-Grady, C. Elias, G. R. Fanning, A. G. Steigerwalt, C. M. O'Hara, G. K. Morris, P. B. Smith, and D. J. Brenner. 1985. Biochemical identification of new species and biogroups of Enterobacteriaceae isolated from clinical specimens. J. Clin. Microbiol. 21:46-76. Hawkey, P. M. 1984. Providencia stuartii: a review of a multiply antibiotic-resistant bacterium. J. Antimicrob. Chemother. 13: 209-226. Hawkey, P. M., S. J. Pedler, and A. Turner. 1983. Comparative in vitro activity of semisynthetic penicillins against Proteeae. Antimicrob. Agents Chemother. 23:619-621. Hendrickson, D. A. 1985. Culture media, p. 1051-1092. In E. H. Lennette, A. Balows, W. J. Hausler, Jr., and H. J. Shadomy (ed.), Manual of clinical microbiology, 4th ed. American Society for Microbiology, Washington, D.C. Hickman, F. W., J. J. Farmer III, A. G. Steigerwalt, and D. J. Brenner. 1980. Unusual groups of Morganella ("Proteus") morganii isolated from clinical specimens: lysine-positive and ornithine-negative biogroups. J. Clin. Microbiol. 12:88-94. Hickman, F. W., A. G. Steigerwalt, J. J. Farmer III, and D. J. Brenner. 1982. Identification of Proteus penneri sp. nov., formerly known as Proteus vulgaris indole negative or as Proteus vulgaris biogroup 1. J. Clin. Microbiol. 15:1097-1102. Hickman-Brenner, F. W., J. J. Farmer III, A. G. Steigerwalt, and D. J. Brenner. 1983. Providencia rustigianii: a new species in the family Enterobacteriaceae formerly known as Providencia alcalifaciens biogroùp 3. J. Clin. Microbiol. 17:1057-1060. Klastersky, J., A.-M. Bogaerts, J. Noterman, E. Van Laer, D. Daneau, and E. Mouawad. 1974. Infections caused by Providence bacilli. Scand. J. Infect. Dis. 6:153-160. Kocka, F. E., S. Srinivasan, M. Mowjood, and H. S. Kantor. 1980. Nosocomial multiply resistant Providencia stuartii: a long-term outbreak with multiple biotypes and serotypes at one hospital. J. Clin. Microbiol. 11:167-169. 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. McHale, P. J., F. Walker, B. Scully, L. English, and C. T. Keane. 1981. Providencia stuartii infections: a review of 117 cases over an eight year period. J. Hosp. Infect. 2:155-165. Penner, J. L., N. A. Hinton, and J. N. Hennessy. 1975. Biotypes of Proteus rettgeri. J. Clin. Microbiol. 1:136-142. Penner, J. L., M. A. Preston, J. N. Hennessy, L. J. Barton, and M. M. Goodbody. 1982. Species differences in susceptibilities of Proteeae spp. to six cephalosporins and three aminoglycosides. Antimicrob. Agents Chemother. 22:218-221. Phillips, E., and P. Nash. 1985. Culture media, p. 1051-1092. In E. H. Lennette, A. Balows, W. J. Hausler, Jr., and H. J. Shadoniy (ed.), Manual of clinical microbiology, 4th ed. American Society for Microbiology, Washington, D.C.

IDENTIFICATION OF P. STUARTII

VOL. 26, 1988 17. Prentice, B., and B. L. Robinson. 1979. A review of Providencia bacteremia in a general hospital, with a comment on patterns of antimicrobial sensitivity and use. Can. Med. Assoc. J. 121:745749.

18. Shlaes, D. M., and C. A. Curnre. 1983. Endemic gentamicin resistance R factors on a spinal cord injury unit. J. Clin. Microbiol. 18:236-241.

327

19. Tomfohrde, K. M., D. L. Rhoden, P. B. Smith, and A. Balows. 1973. Evaluation of the redesigned Enterotube-a system for the identification of Enterobacteriaceae. Apple. Microbiol. 25: 301-310.

20. Warren, J. W. 1986. Providencia stuartii: a common cause of antibiotic-resistant bacteriuria in patients with long-term indwelling catheters. Rev. Infect. Dis. 8:61-67.