A Multicenter Study Evaluating the Current Strategies for Isolating Staphylococcus aureus Strains with Reduced Susceptibility to Glycopeptides

Abstract

Glycopeptide-intermediate Staphylococcus aureus (GISA) and heterogeneous GISA (hGISA) strains are notoriously difficult to detect in the diagnostic laboratory. The clinical importance of GISA, and particularly hGISA, will only be obvious when a definitive detection method is available. A few novel GISA and hGISA detection methods have been proposed; however, their validity has never been tested on a significant scale and in different laboratories. This study compares three screening methods for detecting GISA and hGISA strains in 12 laboratories, using a blind panel of 48 strains with known glycopeptide susceptibilities. The three screening methods used were brain heart infusion agar with 6 mg/liter vancomycin (BHIA6V) (CDC/CLSI), Mueller-Hinton agar with 5 mg/liter teicoplanin (MHA5T) (European Antimicrobial Resistance Surveillance System [EARSS]), and the macrodilution method Etest (MET) (EARSS), with population analysis profile-area under the curve analysis as the gold standard. Sensitivity and specificity were highest for MHA5T and MET, which identified 82.5% and 85.9% of strains, respectively. BHIA6V had poor sensitivity, particularly for hGISA (11.5% of strains were detected), and gave the largest interlaboratory variation in performance. MET exhibited the least interlaboratory variation. It is essential that laboratories use appropriate methods to detect GISA/hGISA strains so that the prevalence and clinical importance of these strains can be assessed properly.

The emergence of glycopeptide resistance in Staphylococcus aureus will have a significant impact on human health. In recent years, glycopeptide-resistant S. aureus (GRSA) and both homogeneous glycopeptide-intermediate S. aureus (GISA) and heterogeneous GISA (hGISA) have increasingly been reported (3, 14, 18). The clinical significance of GRSA and GISA seem to be in little doubt, and there is mounting evidence that heteroresistance is associated with failure of vancomycin therapy (7, 9, 13). However, in order to establish the prevalence and clinical relevance of GISA, and particularly hGISA, a reliable method for their detection must be established.

GRSA strains exhibit vancomycin MICs of ≥32 mg/liter, having acquired vanA from Enterococcus faecalis/Enterococcus faecium (3, 4). Their identification in vitro is assumed to be straightforward using standard protocols; however, their identification with automated systems is reported to be questionable (16). GISA and hGISA isolates, on the other hand, exhibit vancomycin MICs of 4 to 8 and 2 to 4 mg/liter, respectively, and have a mechanism of resistance which has not been defined fully. Accordingly, the detection of GISA, and particularly hGISA, has been beset by problems due to unreliable methodologies (7, 15, 17). Currently, the most reliable method for definitive identification of GISA/hGISA, which has been used in a number of surveillance studies, is the population analysis profile-area under the curve (PAP-AUC) method (2, 8, 19). This method is a modified population analysis method using an analysis protocol and criteria specifically designed to discriminate between glycopeptide-sensitive S. aureus (GSSA), hGISA, and GISA (19). Unfortunately, this method is labor-intensive, and its performance in a diagnostic laboratory with large numbers of strains is untenable. Thus, a reliable screening agar would be preferable in terms of ease of use and cost. Several GISA/hGISA screening methods have been proposed, including the use of various screening agars (1, 5, 6) or interaction agars (H. Hanaki, S. Ohkawa, Y. Inaba, T. Hashimoto, and K. Hiramatsu, Abstr. 38th Intersci. Conf. Antimicrob. Agents Chemother, abstr. C132, 1998) and population studies (5). However, these have not been assessed for comparability and interlaboratory consistency. This study seeks to address this problem by examining three of the most commonly used and recommended methodologies in a multicenter, multinational comparison, using the PAP-AUC method as the gold standard (19).

MATERIALS AND METHODS

Study design.

This study was designed to evaluate the three most common identification methods for detecting Staphylococcus aureus isolates with reduced susceptibility to glycopeptides, including a vancomycin screening agar, as recommended by CDC and CLSI (www.cdc.gov/ncidod/dhqp/ar_visavrsa_labFAQ.html); a teicoplanin screening agar, as recommended by the European Antimicrobial Resistance Surveillance System (http://www.rivm.nl/earss/Images/Earss%20manual2005_tcm61-21261.pdf); and the macrodilution method Etest (MET) (EAS 003; AB Biodisk). Twelve laboratories participated, representing geographically diverse regions throughout the world, with two in the United Kingdom, three in the United States, two in Belgium, one in Switzerland, one in Sweden, one in France, one in Australia, and one in Russia. Each laboratory was furnished with 48 strains of known glycopeptide susceptibility (laboratories were blinded to these data), protocols for three methods, and all media and antimicrobials necessary to complete the study. The strain set included three control strains, namely, Mu50, Mu3, and ATCC 29213, and triplicates of clonally distinct clinical strains representing five GISA strains, five hGISA strains, and five GSSA strains (Table 1). All GISA strains exhibited vancomycin MICs of 8 mg/liter or more by standard CLSI methods and a PAP-AUC value denoting GISA, i.e., ≥1.3 (19). hGISA strains exhibited vancomycin MICs of 1.5 to 4 mg/liter by standard CLSI methods and a PAP-AUC value denoting hGISA, i.e., 0.9 to 1.29 (19). GSSA strains exhibited vancomycin MICs of <2 mg/liter by standard CLSI methods and a PAP-AUC value denoting vancomycin susceptibility, i.e., <0.9 (19). The PAP-AUC method uses overnight cultures of a test organism to inoculate, via a spiral plater, a range of plates containing 0 to 8 mg/liter vancomycin. After 48 h of incubation, the resulting viable counts are plotted against the vancomycin concentration, and the area under the curve is compared to that for the known heterogenous vancomycin-intermediate S. aureus strain Mu3 (18, 19). A results sheet was included to detail how results should be noted.

TABLE 1.

Strain identities and origins

Strain identifier	PAP-AUC value	Reference or source
GISA strains
3759	1.57	10
PC3	1.4	12
LIM3	1.43	11
Michigan	1.9	14
New Jersey	1.8	14
hGISA strains
Duf France	1.16	Evelyne Lecaillon, Community Hospital, Perpignan, France (personal communication)
AG Liverpool	1.29	John Corkill, Royal Liverpool and University Hospital, United Kingdom (personal communication)
LIM1	1.14	9
23 Southampton	1.04	Andy Tuck, PHL, Southampton General Hospital, United Kingdom (personal communication)
Sweden MDRSA79	1.21	Ann Bolmström, Sweden (personal communication)

Laboratory methods. (i) BHIA6V screening agar.

In each participating laboratory, brain heart infusion agar (BHIA; BBL, Becton Dickinson, MD) plates were manufactured to contain 6 mg/liter vancomycin (Eli Lilly, Basingstoke, United Kingdom) (BHIA6V). Strains were subcultured from swabs onto blood agar, and after overnight growth, several colonies were suspended in 0.9% saline to obtain an inoculum with equivalent turbidity to a McFarland standard of 0.5. Ten microliters of inoculum was delivered onto the surface of the agar, and the plate was incubated at 35°C in air for 24 and 48 h. Growth of more than one colony was recorded at both 24 and 48 h. A strain was considered positive if growth of two or more colonies occurred after 24 h.

(ii) MHA5T screening agar.

In each participating laboratory, Mueller-Hinton agar (MHA; Oxoid, Basingstoke, United Kingdom) plates were manufactured to contain 5 mg/liter teicoplanin (Merrell Dow Pharmaceuticals, Staines, United Kingdom) (MHA5T). Strains were subcultured from swabs onto blood agar, and several colonies were suspended in 0.9% saline to obtain an inoculum with equivalent turbidity to a McFarland standard of 2. Ten microliters of inoculum was delivered onto the surface of the agar, and the plate was incubated at 35°C in air for 24 to 48 h. Growth of more than one colony was recorded at both 24 and 48 h. A strain was considered positive if growth of one or more colonies occurred after 48 h.

(iii) MET.

MET was performed according to the instructions in the manufacturer's manual (EAS 003; AB Biodisk). Briefly, several colonies were suspended in Mueller-Hinton broth (Oxoid, Basingstoke, United Kingdom) to obtain an inoculum equivalent to a 2 McFarland standard. One hundred microliters of inoculum was evenly streaked onto a 90-mm BHIA (BBL, Becton Dickinson, MD) plate and allowed to dry. Both teicoplanin and vancomycin Etest strips (AB Biodisk, Solna, Sweden) were applied to the surface of the agar, and the plates were incubated at 35°C in air for 24 and 48 h. Zones were read at complete inhibition, with care, to visualize hazy growth and microcolonies. A strain was considered positive if readings were ≥8 mg/liter for vancomycin and teicoplanin or ≥12 mg/liter for teicoplanin alone.

Analysis.

Each laboratory returned results in the form of a datasheet detailing the screening method used, the number of colonies grown at 24 h, the number of colonies grown at 48 h, the vancomycin Etest reading, and the teicoplanin Etest reading. For each laboratory and method, the 48 strains were designated phenotypes dependent upon screening agar criteria, which were then compared to the original phenotypes determined by the PAP-AUC method (19) (Table 1). Because the three methods used are not capable of distinguishing hGISA from GISA strains, both phenotypes were classified as glycopeptide intermediate (GI). Percentages of correctly identified glycopeptide-susceptible (GS) and glycopeptide-intermediate (GI) strains were calculated along with the method sensitivity, specificity, and positive and negative predictive values. Sensitivity refers to how good a method is at identifying GI strains, specificity refers to how good a method is at identifying GS strains, the positive predictive value refers to the probability that a positive result is correct, and the negative predictive value refers to the probability that a negative result is correct.

RESULTS

The percentages of strains correctly identified are detailed in Table 2. Among 16 GS strains in the set, BHIA6V, MHA5T, and MET correctly identified means of 15.58, 12.17, and 14.08 strains, respectively. However, among a possible 32 GI strains (16 GISA and 16 hGISA strains), BHIA6V, MHA5T, and MET were able to correctly identify means of 11.25, 27.58, and 26.83 strains, respectively. The mean total percentages of strains correctly identified by BHIA6V, MHA5T, and MET were 55.9%, 82.5%, and 85.9%, respectively.

TABLE 2.

Percentages of correctly categorized strains for all three screening methods in all laboratories

Screening method	% Correctly identified strains (mean ± SD)
	GS strainsa	hGISA strains	GISA strains	hGISA and GISA strains	Total
BHIA6Vb	97.4 ± 7.3	11.5 ± 25.4	58.9 ± 29.7	35.2 ± 23.9	55.9 ± 14.59
MHA5Tc	76.0 ± 25.5	78.7 ± 20.4	93.8 ± 14.3	86.2 ± 15.9	82.8 ± 9.93
METd	88.0 ± 9.4	70.3 ± 23.3	97.4 ± 5.6	83.8 ± 13.2	85.9 ± 8.44

^aGS, glycopeptide susceptible.

^bScreening agar using brain heart infusion agar plus 6 mg/liter vancomycin.

^cScreening agar using Mueller-Hinton agar plus 5 mg/liter teicoplanin.

^dMacrodilution Etest.

The standard deviation (SD) was calculated to determine the variation between laboratories in results for each method. The mean SD was high for BHIA6V, at 14.59, and lower for MHA5T (9.93) and MET (8.44) (Fig. 1). The percentages of false-positive results identified by BHIA6V, MHA5T, and MET were 0.42%, 3.8%, and 1.83%, respectively, while the percentages of false-negative results were 20.66%, 4.42%, and 5.25%, respectively.

FIG. 1.

Variation in performances of BHIA6V, MHA5T, and MET in 12 laboratories.

The means for correctly identifying GI strains can be separated further into values for GISA and hGISA strains. MHA5T and MET correctly identified 93.8% and 97.4% of GISA strains, respectively, and 78.7% and 70.3% of hGISA strains, respectively, while BHIA6V correctly identified only 58.8% of GISA strains and only 11.5% of hGISA strains.

In comparing how well each strain was identified in all methods with its PAP-AUC value (as a measure of the level of glycopeptide resistance) (Fig. 2), it was observed that BHIA6V correctly identified only strains with high PAP-AUC values (usually GISA strains [>1.5]), whereas both MHA5T and MET appeared to correctly identify GISA and hGISA strains with PAP-AUC values of >1.

FIG. 2.

Comparison of methods in relation to glycopeptide susceptibility, as determined by the PAP-AUC method. Full details of the PAP-AUC protocol and criteria can be found in Materials and Methods.

In an attempt to optimize the two screening agars, we amended the criteria used to determine positive results. For BHIA6V, the criterion for a positive test result was changed from growth of two or more colonies after 24 h of incubation to growth of one or more colonies after 48 h. For MHA5T, the criterion was changed from growth of one or more colonies after 48 h of incubation to growth of two or more colonies after 48 h. Changing the criteria improved the mean total percentages of correctly identified strains by 6.6% and 1.22% for BHIA6V and MHA5T, respectively. No criterion amendments, including the lowering or raising of positive threshold readings, could be found to increase the numbers of strains correctly identified by MET.

Mean sensitivities for BHIA6V, MHA5T, and MET are shown in Table 3. BHIA6V shows low mean sensitivity and negative predictive values but relatively high specificity and positive predictive values. MHA5T and MET show considerably higher sensitivity and negative predictive values but slightly lower specificity and positive predictive values. Overall, sensitivities, specificities, and predictive values correlate with the analysis of percentages of strains correctly identified.

TABLE 3.

Sensitivity, specificity, and positive and negative predictive values for all three screening methods in all laboratories

Screening method	Mean sensitivity (%)			Mean specificity (%)	Mean positive predictive value (%)			Mean negative predictive value (%) for all GI strains
	Total GI strainsa	GISA strains	hGISA strains		Total GI strains	GISA strains	hGISA strains
BHIA6V	35.16	58	11.47	97.4	98.43	97.73	21.37	45.15
MHA5T	85.95	92.21	79.71	75.55	82.2	81.76	79.89	79.09
MET	82.04	94.28	69.3	89.09	93.98	90.05	87.19	74.42

^aGI, glycopeptide intermediate.

DISCUSSION

This study evaluates the performances of three screening methods for detecting S. aureus strains with reduced susceptibility to glycopeptides. These methods were tested in 12 laboratories throughout the world, using identical media, antimicrobials, and isolates. Analysis of data from all laboratories shows that BHIA6V performed least effectively, with <60% of isolates correctly identified and sensitivity and specificity values of 35.16% and 97.4%, respectively. BHIA6V also produced the greatest number of false-negative results and showed the greatest variation between laboratories, with percentages of correctly identified strains ranging from 37.5% to 89.6%. A high degree of variability when using BHIA6V produced in-house was previously reported, but this study confirms this finding (16).

Overall, the total percentages of strains correctly identified and sensitivity and specificity values were similar for MHA5T and MET. However, if the criterion used to identify positive results by MHA5T was increased to growth of two or more colonies at 48 h, instead of one or more colonies, then the mean total percentage of strains correctly identified increased to 84.03% (from 82.9%), compared to 85.9% for MET. This increase in correct identification occurred in 6 of 12 laboratories, with the other 6 laboratories showing no change in numbers of correctly identified strains. MHA5T had slightly fewer false-negative results but twice as many false-positive results as MET, although the percentages are small (1.8% and 3.8%, respectively). Although the sensitivity and specificity of MHA5T are slightly better at detecting GISA, especially hGISA (10% higher), than those of MET, it is at the cost of falsely identifying GSSA as GI. Again, as with the numbers for false-positive and -negative results, the mean positive and negative predictive values for MHA5T and MET indicate that MHA5T is slightly better at predicting glycopeptide susceptibility but less accurate at predicting intermediate glycopeptide resistance.

Although these data show similar performances for MHA5T and MET, overall it seems that MET has the advantage of producing fewer false-positive results. The cost of introducing any screening test to a diagnostic laboratory is important, and both BHIA6V and MHA5T are relatively low in cost, while MET has a greater impact on laboratory finances. However, laboratories should also be aware that any positive strain detected by a screening method would undergo confirmatory testing. This would incur additional costs, and hence any reduction in the false-positive rate would reduce unnecessary costs. Diagnostic laboratories must assess the relative benefits of the MET and MHA5T methods with respect to initial screening costs and the costs of confirmatory testing. MET also shows less variation in performance between laboratories, a significant factor when considering diagnostic methods.

The evaluation of screening methods for detecting GISA and hGISA strains is of significant value if we are to fully understand the prevalence and, hence, clinical importance of GISA, particularly hGISA. Furthermore, such investigations are crucial if data on the efficacy of new drugs, such as linezolid and daptomycin, directed against methicillin-resistant S. aureus strains with reduced susceptibility to glycopeptides are going to be at all meaningful.