Drug susceptibility testing of clinical isolates of streptococci and enterococci by the Phoenix automated microbiology system

Background Drug resistance is an emerging problem among streptococcal and enterococcal species. Automated diagnostic systems for species identification and antimicrobial susceptibility testing (AST) have become recently available. We evaluated drug susceptibility of clinical isolates of streptococci and enterococci using the recent Phoenix system (BD, Sparks, MD). Diagnostic tools included the new SMIC/ID-2 panel for streptococci, and the PMIC/ID-14 for enterococci. Two-hundred and fifty isolates have been investigated: β-hemolytic streptococci (n = 65), Streptococcus pneumoniae (n = 50), viridans group streptococci (n = 32), Enterococcus faecium (n = 40), Enterococcus faecalis (n = 43), other catalase-negative cocci (n = 20). When needed, species ID was determined using molecular methods. Test bacterial strains were chosen among those carrying clinically-relevant resistance determinants (penicillin, macrolides, fluoroquinolones, glycopeptides). AST results of the Phoenix system were compared to minimal inhibitory concentration (MIC) values measured by the Etest method (AB Biodisk, Solna, Sweden). Results Streptococci: essential agreement (EA) and categorical agreement (CA) were 91.9% and 98.8%, respectively. Major (ME) and minor errors (mE) accounted for 0.1% and 1.1% of isolates, respectively. No very major errors (VME) were produced. Enterococci: EA was 97%, CA 96%. Small numbers of VME (0.9%), ME (1.4%) and mE (2.8%) were obtained. Overall, EA and CA rates for most drugs were above 90% for both genera. A few VME were found: a) teicoplanin and high-level streptomycin for E. faecalis, b) high-level gentamicin for E. faecium. The mean time to results (± SD) was 11.8 ± 0.9 h, with minor differences between streptococci and enterococci. Conclusion The Phoenix system emerged as an effective tool for quantitative AST. Panels based on dilution tests provided rapid and accurate MIC values with regard to clinically-relevant streptococcal and enterococcal species.


Background
Resistance of streptococci and enterococci to different classes of antimicrobials is increasing worldwide [1,2]. Reduced susceptibility or resistance to penicillin and other β-lactams have been reported with particular frequency in the case of Streptococcus pneumoniae and viridans streptococci [1,3]. Resistance to macrolides and fluoroquinolones has been described in viridans and βhemolytic streptococci [4,5]. Notably, glycopeptide resistance is an increasing threat among enterococcal species, especially Enterococcus faecium and E. faecalis [2].
Since clinical isolates of all the above species may produce severe infections (sepsis, meningitis, endocarditis), rapid and accurate identification (ID) and antimicrobial susceptibility testing (AST) are of great significance. For these bacteria, evaluation of drug susceptibility using disk diffusion methods can be problematic since the technique does not provide quantitative values. In this context, the Clinical Laboratory Standards Institute (CLSI; formerly NCCLS) established that S. pneumoniae isolates should not be reported as penicillin-resistant or intermediate based solely on an oxacillin zone ≤ 19 mm [6]. For these strains, minimum inhibitory concentration (MIC) values of penicillin and cefotaxime/ceftriaxone or meropenem are mandatory. Correct detection of vancomycin resistance among enterococci is also critical, since diffusion is difficult for large molecules and interpretative criteria are based on limited differences in the inhibition zone. MIC values may be of further assistance in discriminating enterococcal species carrying different van-type determinants.
Reproducibility of results, accuracy of species ID, precision of quantitative AST results, turnaround time, availability of data for epidemiological monitoring, and costeffectiveness represent compelling reasons that support the choice of an automated system for clinical bacteriology. Systems of this type are now available for catalasenegative gram-positive cocci [7][8][9]. In the year 2000, Becton Dickinson (BD Diagnostic Systems, Sparks, MD) introduced the Phoenix Automated Microbiology System for ID and AST of gram-negative bacilli, staphylococci, and enterococci [10][11][12][13]. More recently, a panel for streptococcal species (SMIC/ID-2 panel) has been added [14,15].
The present study was designed to evaluate the performance of the Phoenix system for the quantitative determination of drug susceptibility of streptococcal and enterococcal isolates carrying different resistance determinants. Time to results was also analyzed.

Results
Bacterial isolates are listed in Table 1. AST results are expressed as raw data (i.e., as results not interpreted by the expert system). For streptococcal isolates (Table 2), the overall essential agreement (EA; 91.9%) and categorical agreement (CA; 98.8%) between the Phoenix system and the Etest method were satisfactory. No VME were produced. Very low rates of major errors (ME; 0.1%) and minor errors (mE; 1.1%) were found. For enterococcal isolates (Table 3), the overall EA and CA were 97% and 96%, respectively. Small numbers of errors were found: VME (0.9%), ME (1.4%), mE (2.8%).
As shown in Tables 2 and 3, the overall performance for different antimicrobials was rather satisfactory (mean EA for streptococci ≥ 91%, for enterococci ≥ 97%). Poor performances were, however, noticed for streptococci in the case of clindamycin (EA, 67.3%) and erythromycin (EA, 88.6%), and for enterococci in the case of teicoplanin (EA, 87%). CA rates ≥ 97% and ≥ 94% were obtained for streptococci and enterococci, respectively. However, the low CA rate obtained in the case of linezolid for enterococci (89%) should be noticed.   Table 1) are not presented in Table 5. For these species, mean CA value was 88.9%; mE and ME were 7.2% and 5.3%, respectively. Errors were restricted to chloramphenicol, tetracycline, linezolid, and quinupristin-dalfopristin.
Streptococcal and enterococcal isolates characterized by a wide spectrum of drug susceptibility have been investigated to assess the performance of the Phoenix system using both the new SMIC/ID-2 panel for streptococci and the established PMIC/ID-14 panel for enterococci. Both panels allowed testing clinically-relevant drugs: penicil-  For streptococci, results in agreement with ours have been obtained by Japanese investigators [14,15] that used SMIC/ID panels differing from those used by us. As compared to the automated Vitek system, the Phoenix appeared more accurate in determining MIC values for pneumococci [18]. Excellent results were obtained with both systems with regard to S. agalactiae [9]. Among enterococci, our results showed lower performances with respect to what has been reported by others using both the Phoenix and the Vitek systems [9,12,19]. In this respect, it should be noted that a high proportion (46/100) of our isolates carried glycopeptide resistance determinants and that 17/100 isolates were of species other than E. faecalis and E. faecium. Problems with AST results for the above species have also been reported by others using automated systems [20].
Unsatisfactory results have been obtained for clindamycin in the case of streptococci (EA, 67.3%): MIC values produced by the Phoenix were, in fact, usually lower than those obtained with the Etest. Investigated strains, however, showed MICs consistently within breakpoint values, thus limiting the clinical impact of the error. The reduced   response to clindamycin was not related to the expression of resistance determinants (inducible/constitutive MLS B phenotype, or efflux; data not shown). Overall, small numbers of VME and ME were obtained. Most errors were classified as mE, due to minor differences in MIC values around breakpoint concentrations. Regarding ten enterococcal isolates, the Phoenix system gave MIC values of 4 µg/ml for linezolid, thus classifying the isolates as "intermediate". MIC values given by the Etest were instead of 2 µg/ml (i.e., "susceptible"). The finding appears to reflect a problem already noticed by Phoenix users with regard to this drug. Thus, increased linezolid MICs need to be verified using a reference method. Regarding penicillin, mE were encountered in 6/150 streptococci. In the latter cases, MICs were overestimated by the Phoenix system. Overall, the results of the present study were superior to those previously reported for both the Phoenix and the VITEK 2 system [14,15,18].
A mE rate < 10% is recommended for accepting AST results, and up to 3.0% ME and 1.5% VME are acceptable [21]. Overall, the results produced by the Phoenix were well within the recommended range. However, an ME rate of 5.3% was obtained for enterococci other than E. faecium and E. faecalis.

Conclusion
The Phoenix system appears an effective diagnostic tool for infections caused by streptococci and enterococci. Panels based on dilution tests were able to deliver adequate quantitative AST results in short times. This is especially relevant for severe infections caused by aggressive species such as S. pyogenes, S. pneumoniae, E. faecalis, and E. faecium. For these organisms, interpretation of susceptibility data varies according to the infection site and -in the case of large molecules that hardly diffuse in solid media (e.g., glycopeptides) -results obtained using dilution methods are to be preferred [20,22].

Clinical isolates
Bacterial strains were obtained from routine clinical specimens at the Microbiology Laboratory of the Ospedale di Circolo, Varese (Italy). A total of 250 non-duplicate clinical isolates of gram-positive, catalase-negative cocci were studied. When needed, species ID was obtained by molecular methods [23]. The following species were investigated (Table 1) [24]. Isolates were stored at -70°C in Todd-Hewitt broth containing 20% glycerol. Before performing ID and AST assays, the isolates were passed twice on Mueller-Hinton agar containing 5% sheep blood (Oxoid SpA, Milan, Italy) to get them to an active growth stage following metabolic inactivity while frozen.

Analysis of data and discrepancies
AST results are reported as raw data, i.e. without interpretation by the expert system. Results have been categorized as susceptible (S), intermediate (I), or resistant (R) according to CLSI [6]. MIC values produced by the Phoenix method (TEST) have been compared to those obtained with the Etest method (COMP). The following definitions have been adopted: 1) EA (i.e., MIC values of TEST panel equal to or within ± 1 dilution of the COMP value); 2) CA (i.e., TEST and COMP MIC values agree using insterpretative CLSI criteria); 3) mE (i.e., COMP is S or R and TEST is I; alternatively, COMP is I and TEST is S or R); 4) ME (i.e., COMP is S and TEST is R; the percentage of major errors has been calculated only for susceptible isolates); 5) VME (i.e., COMP is R and TEST is S; the percentage of very major errors has been calculated only for resistant isolates). In the case of ME or VME, the organism has been re-tested with both the Phoenix and the Etest. When differences persisted, the results have been considered discordant.

Quality controls
As controls, the following strains have been included in each run: S. pneumoniae ATCC 49619, S. agalactiae ATCC 13813, and E. faecalis ATCC 29212.