The ~erfQr?na?

2~e
of the Vitek s~~tcttv
nvrd %heAPP 20s p
to the species level was eva~~~ute~
in
cov?~~ariso~with a ~o}~ve~#io~~al
reference test scheme. A total
of 369 isolates originally derived from a sample from 97 medical centers were tested, and the ~er~e?~tageof correctly icientified organisms was ~VitekiA~~~:97.6%179.2%, 95.3%/91.2%,
and $6.0%/20.7% forEnterococcus faecalis, Enterococcus

j~e~%~~~~ng
enterococci

erococcal infections

Nosocomial
ecau5e

e~terococca~ i

agents. Furthermore, with certain i

especially bloodstream infections
endocarditis, more precise
tant because of naturally occ
antimicrobial susceptibility among enterococcal
strains and species. Another emerging factor e
was to evaluate
?&medical
Micrabiology Division, Department of
Pathology, University of Iowa College of Medicine, Iowa City,
Iowa, USA.
Address reprintre9uesfs to Dr. H.S. Sader, Disciglina de
Doencas Jnfecciosas e Farasitirias, Unversidade Federal de SBo
Paula, Escola Paulista de Medicine, Rua Botucatu, 740 CEP
04023-062, Sio Paula, SP, 5razil.
Received 15 June 1995; revised and accepted 7 July 1995.

DIAGN MICRQBlOL INFECT DllS 1995;22:315-319

0 1995 H.S. Sader et al.

analyzed 369

nOsoc0

0732-8893195
SSDI 0732-8893(95)00146-2

isolates consecutively collected in the lastquarter of

1992. These strains were isolated from bloodstream
or other nonurinary infections in 97 U.S. medical
centers distributed in 46 states an
Columbia (Jones et al., 1995). The isolates were
identified locally, at the time the OrganiSmS Were
isolated,using the API 205 System. A.lI Centers received products from the same lot and followed the
same manufacturer-provided
instructions to perform the test. The selection of the 369 isolates evaluated in the present study was based on the initial
API results as follows: 104 isolates ide~ti~ed as E.
faecttlis(from $3 medical centers) and all vi
@&is isolates available (at least one from ea
ter, n = 265) were retested by the Vit
positive identification
(GM) card us
industrial mode software a 8.1. The Vite
cation was supplemented with motil
instrument report
en species that
tion (e.g., E. fuecium and E. gallimum or E. cassehflavus) (Facklam and Collins, 1989).
The conventional test scheme (CTS) for enterococcal species identification was considered to be
the gold standard to assess the accuracy and errors of the commercial systems. If the commercial
systems agreed with each other and both showed
high confidence (excellent identification by API and
>95% confidence by Vitek), both were considered
correct and CTS was not performed. Although a
concordance between two commercial systems does
not necessarily indicate an accurate identification,
we believe that the probability of both systems
identifying the same isolate in the same way
with high probability is very low (Buschelman e
1993). Any discrepancies or lo
were further tested by the CTS.
was performed in 37 isolates (33 E. faecim, three E.
fimiis, and one E. guhwm)
that the two commercial systems agreed to each other. Absolute agreement was obtained between methods (P c ,001).
The CTS was performed by using a modified version (Buschelman et al., 1993) of the method described by Facklam and Collins (1989) on 117 isolates
demonstrating intermethod discord. Briefly, carbohydrate fermentation was tested in brain heart infusion (BHI) broth base with 1% arabinose, lactose,
1, raffinose, sorbitol, sorbose, and sucrose
icrobiologicals, Tualatin, OR, USA). Deam3Mhm Of arginine was also tested in BHI broth, and
ity was determined using motility test me. Pigment production was assessed by examining the growth on a cotton swab.
The performance of the commercial systems was
evaluated based on two parameters: accuracy and
Predictive value (pv). The first parar;leter indicates

The ability of the Vitek GPI card to correctly i

ci isolates to the s

the API 20s system

erroneous species i

however, only six E. auium isolates were correctly

identified (major error of 69%). The API 20s usually
identified NFNF isolates as either E. faecalis or E.
faecium (Table 1). On the other hand, the Vitek System correctly identified 25 (86%) isolates, and only
three isolates were misidentified (major error of
10.3%). The misidentifications occurred with E. ra@nosus and E. solitarius, species that are not included
in the software data
shown in other studi
et al., 1991), both E.
fied as E. avim (<87% confidence; marginal separation). The E. soliturius isolate was misidentified as
E. fiecalis. These two s
era1 phenotypic characteristics, making differentiation very difficult (Ruoff et al., 1990; Facklam and
Collins, 1989). In addition, E. sditaariushas been removed from the Enferococcus genus (Williams et al.,

Species identification based on the c~~ve~~~~~ lest scheme when the ~5rnmer~~a~
systems did no%agree wit
*Questionable pattern (61% E. fieciurn, 22% E. faecalis).

It was considered a major error when the system provided a wrong identification, and minor
error when the system did not provide a species identification.
WFNF, non-faecalis-non-fgecium species.

system is. The analysis oi the kV results for the

Vitek demmstrated that the answer provided by
this system is very reliable. Based on the data gen-

319

Predictive Values of Commercial

Systems According to Species
Predictive Value

(S/o)
Species

ViEek

AM 20s
95.2
a.8
54.5

A cost analysis, not includin

GP1 card costs between $3.50 a
depending
on the vsler

correctly identified as E. faecdis by the API

WOUld
205 but misidentified by the Vitek System. HOWisolates initially identified as E.
ever, among the
OS,
only one was miside~~tified
@&is by the A
by the Vitek System (Table l), and the isolate was
identified as E. fu~cium with HOWCC&
1% E.
faecium, 22% E, faecalis: questionablie
The FV results should be analyzed with caution,
because the species distribution of the collection is
atypical-it has a very low percentage of E. faecalis
isolates compared with that commonly reported in
the literature. If the collection had a higher propsrtion of E. faecalis isolates, the PV value for this species would be higher for both identification systems.
However, the PV for E. faecium would be lower,
especially for the API 2OS system, which identified
18% of the E. fmxdis as E. fnccium. Conversely, the
l?V of the Vitek System for E. fcccitrm would not
change significantly, as < 3% of the E. @calis isolates was misidentified by this product (Table 1).
The PV of the API 20s for E. fuecium showed that
16.2% of isolates identified as E. faecium and approxy half of the isolates identified as one of the
species were found to belong to a different
species (Table 3). The most common API 2OS error
was the identification of E. fmmlis as E, fuecium (TaMe 2). Almost 10% of the isolates initially identified
as E. fnecium by the API 20s were found to be E.
fkaecdisby CTS. Although the version we evaluated
did not show excellent results for any of the enterococcal species, the API 20s was modified following
this study (Jones et al., 1995), and it may have been
improved. In addition, the comparison of the two
methods was biased by the fact the API 20s tests
were performed by each center, whereas all of the
Vikk System tests were performed by a single investigator laboratory.

Buschelrnan BJ, Bale MJ, Jones RN (1993) Species iden&

fication and determination of high-level resistance
among enterococci: comparison study of sterile body

single plate.
Qur f~~d~gs confirm and extend tide results of
other studies (Buschcl an et al., 1993; Buoff et al.,
1990; Willey et
1993) in showing the excellent
performance of e Vitek System GPI card (version
3 8.1) in identi
faedis asd E. fmd
ties. However,
tificatisn of NFNF
was also proven to be acceptable, which c
with earlier investigations. VJe were able to correctly
identify 90% of NFNF isolates, with a predictive
value of 89%. Although E. f~emlis and @. faecium
ide~~ficatio~s are responsible for the last majority
of clinical infections, there have been several reports
of NFNF isoiates causing significant diseases (Pate1
et al., 1993; Jan Goethem et al., 1994). In addition,
some NFNF species may be more intrinsically resistant to antimicrobial agents (glycopeptides,
lins, etc.) than E. faecutis or E. faecitm (Gordon et al.,
1992). The use of a newer version of the industrial
mode software and two simple ddditianal tests (motility and pigment) can provide very satisfactory accuracy and PV for most NFNF species. The u:3e of
these two tests to complement the identification can
also increase the PV for E. fuecalis and E. faecium,
because NFNF isolates were most commonly misidentified as one of the more frequently isolated
species, especially E. fuecium (Buschelman et al.,
1993; Willey et al., 1993). These two tests are
very inexpensive and can be
the identification confidence is low (<$7%), which
occurred in only 6% of the isolates tested in our
study.

f!uid isolates, 19854991. Diagn Microbial InfecCDis 16:

119-122.
Facklam RR, Collins &ID (1989) Identification of Enterococ-