Diagnostic Accuracy Assessment of Sensititre and Agar Disk Diffusion for Determining Antimicrobial Resistance Profiles of Bovine Clinical Mastitis Pathogens

Abstract

Determining the accuracy and precision of a measuring instrument is pertinent in antimicrobial susceptibility testing. This study was conducted to predict the diagnostic accuracy of the Sensititre MIC mastitis panel (Sensititre) and agar disk diffusion (ADD) method with reference to the manual broth microdilution test method for antimicrobial resistance profiling of Escherichia coli (n = 156), Staphylococcus aureus (n = 154), streptococcal (n = 116), and enterococcal (n = 31) bovine clinical mastitis isolates. The activities of ampicillin, ceftiofur, cephalothin, erythromycin, oxacillin, penicillin, the penicillin-novobiocin combination, pirlimycin, and tetracycline were tested against the isolates. Diagnostic accuracy was determined by estimating the area under the receiver operating characteristic curve; intertest essential and categorical agreements were determined as well. Sensititre and the ADD method demonstrated moderate to highly accurate (71 to 99%) and moderate to perfect (71 to 100%) predictive accuracies for 74 and 76% of the isolate-antimicrobial MIC combinations, respectively. However, the diagnostic accuracy was low for S. aureus-ceftiofur/oxacillin combinations and other streptococcus-ampicillin combinations by either testing method. Essential agreement between Sensititre automatic MIC readings and MIC readings obtained by the broth microdilution test method was 87%. Essential agreement between Sensititre automatic and manual MIC reading methods was 97%. Furthermore, the ADD test method and Sensititre MIC method exhibited 92 and 91% categorical agreement (sensitive, intermediate, resistant) of results, respectively, compared with the reference method. However, both methods demonstrated lower agreement for E. coli-ampicillin/cephalothin combinations than for Gram-positive isolates. In conclusion, the Sensititre and ADD methods had moderate to high diagnostic accuracy and very good essential and categorical agreement for most udder pathogen-antimicrobial combinations and can be readily employed in veterinary diagnostic laboratories.

INTRODUCTION

Antimicrobial therapy is generally the most common way of treating mastitis in dairy cattle (19). Unfortunately, despite the use of the best possible antimicrobial treatments, bacteriological cure rates (e.g., of Staphylococcus aureus mastitis) seldom exceed 50%. Antimicrobial resistance (AMR) is potentially one of the reasons for treatment failures (1); hence, antimicrobial susceptibility testing of udder pathogens is an important step in defining appropriate farm-level treatment protocols.

The most common method used for AMR profiling of bacterial isolates is the agar disk diffusion (ADD) method of Bauer et al. (3). The ADD method has long been used in veterinary diagnostic microbiology due to easy use, low cost, interlaboratory repeatability, and flexibility in the type and number of drugs that can be tested (28). This test has extensively been used for ascertaining antibiograms of bovine mastitis pathogens (18, 22). However, the ADD method is sensitive to changes in operator techniques and zone of inhibition diameter interpretation, and only qualitative results indicating sensitive, intermediate, and resistant are obtained. Therefore, to relate these qualitative results to time-varying concentrations of antimicrobials at the site of infection, quantitative results in the form of MICs were needed (31). In order to speed up the process of MIC determination, various commercial automated MIC susceptibility systems have been developed. One of the commercial in vitro broth microdilution methods used in veterinary microbiological diagnostics for AMR profiling is the Sensititre system (Trek Diagnostic Systems, Cleveland, OH) (27). Results can be determined using either an automated or a manual reading system and are referenced with the CLSI standards (2, 6, 9). The Sensititre MIC testing system is of particular interest compared to other commercial MIC systems because this system offers a MIC panel specifically for bovine mastitis pathogens.

In human medicine, many studies have determined diagnostic agreement between Sensititre and the manual broth microdilution test method with stock organisms and human clinical isolates for assessing intra- and interlaboratory variations in antimicrobial susceptibility testing (9, 12, 16). In veterinary medicine, although many diagnostic laboratories are using commercial antimicrobial susceptibility systems, there is a dearth of validation study data in this regard (30). Papp and Muckle (23) compared a commercial microdilution MIC system (Sceptor system) with an agar dilution method for veterinary clinical isolates. Intertest MIC comparisons were done for Gram-positive and Gram-negative isolates. However, common Gram-negative bovine mastitis pathogens, such as Escherichia coli, were not tested, and the animal sources of these veterinary clinical isolates were also not described. Watson et al. (29) compared a veterinary breakpoint MIC system with the ADD method with common veterinary pathogens. In this study, only a single concentration of various antimicrobials was used, and isolates from bovine mastitis samples were not included. Franklin and Wierup (7) compared the Sensititre MIC system to the agar dilution method for antimicrobial resistance profiling of veterinary pathogens isolated from different animals; however, the intertest MIC comparisons were made on the genus level and the animal sources of isolates were not identified. To date, the Sensititre system has not been compared to a reference broth microdilution MIC test method with bovine clinical mastitis pathogens.

Similarly, other studies involving comparisons between MIC susceptibility systems and ADD methods for AMR profiling of veterinary pathogens are limited in scope to a few udder pathogens and a few antimicrobial drugs used for control and treatment of mastitis (13, 17, 25). These studies did not use the commercial Sensititre system. Furthermore, the Sensititre automatic reading method has not been compared with manual reading methods in the studies involving veterinary pathogens.

The objectives of this study were therefore (i) to predict the diagnostic accuracy of the Sensititre MIC mastitis panel and agar disk diffusion method using the manual broth microdilution MIC test method as the reference, (ii) to assess diagnostic agreement between the agar disk diffusion and manual broth microdilution MIC test methods, (iii) to assess MIC diagnostic agreement between the Sensititre system and the manual broth microdilution MIC test method, and (iv) to assess agreement between the Sensititre automatic reading and manual reading test methods in determining AMR profiles of clinical bovine mastitis pathogens.

MATERIALS AND METHODS

Herd selection, sampling, and bacterial culturing.

Milk samples (n = 3,033) were obtained from quarters of dairy cows with clinical mastitis in 10 provinces across Canada (21). In short, dairy farmers were contacted through local veterinary practitioners or the provincial Canadian Quality Milk Program to submit milk samples from cows with producer-diagnosed clinical mastitis to the Atlantic Veterinary College at Charlottetown, Prince Edward Island, Canada. A total of 1,441 isolates were cultured from these milk samples from 106 dairy farms. Keeping in mind that multiple isolates could be coming from a single farm and that antimicrobial resistance in isolates could potentially be a herd-level factor, it was decided to keep the number of isolates per farm as low as possible for the purpose of statistical independence. Therefore, 457 isolates were selected for comparing the Sensititre system with the ADD method. These isolates were lyophilized and stored afterwards. Two years later, out of these 457 isolates, a random subset (n = 150, with 25 isolates per mastitis pathogen) was selected for validating the Sensititre system and the ADD method using the manual broth microdilution test method as the reference. However, because not all lyophilized samples could be recultured, a total of 119 isolates were tested finally with the manual broth microdilution test method. Bacterial culturing and identification of the organisms in the milk samples were done according to National Mastitis Council guidelines (14). The following reference strains were included in the study: S. aureus ATCC 25923, S. aureus ATCC 29213, Enterococcus faecalis ATCC 29212, Streptococcus pneumoniae ATCC 49619, and Escherichia coli ATCC 25922. Isolates of interest in this study, e.g., S. aureus, Streptococcus uberis, Streptococcus dysgalactiae, E. coli, other streptococci, and Enterococcus spp., were stored in skim milk in a commercial freezer at −20°C.

Antimicrobials.

A Sensititre standard susceptibility mastitis plate (CMV1AMAF) consisting of 10 antimicrobials in serial 2-fold dilutions was used in the study (27). This bovine mastitis plate contains the following antimicrobials: ampicillin, ceftiofur, cephalothin, erythromycin, oxacillin, penicillin, penicillin/novobiocin combination, pirlimycin, sulfadimethoxine, and tetracycline (Table 1). Commercial antimicrobial disks of ampicillin, ceftiofur, cephalothin, erythromycin, oxacillin, penicillin, novobiocin, pirlimycin, and tetracycline were used for the ADD method (Table 1). Since sulfadimethoxine is hardly used for mastitis treatment and control, it was not used for the ADD and manual broth microdilution test methods.

Table 1.

Range of concentrations of antimicrobials used in manual broth microdilution test method, Sensititre bovine mastitis panel, and commercial agar diffusion disks

Antimicrobial	Concn (μg/ml)
	Broth microdilution	Sensititre	Agar diffusion disk
Ampicillina	0.0075–16	0.12–8	10
Ceftiofurb	0.00375–8	0.5–4	30
Cephalothina	0.015–32	2–16	30
Erythromycina	0.00375–8	0.25–4	15
Oxacillina	0.00375–8	2–4	1
Penicillina	0.0075–16	0.12–8	10d
Pirlimycinb	0.00375–8	0.5–4	2
Penicillin-novobiocinb	0.00375–8	1/2–8/16	10d/30
Sulfadimethoxinec		32–256
Tetracyclinea	0.0075–16	1–8	30

^aOxoid Canada, Nepean, Ontario, Canada.

^bBD BBL, Oakville, Ontario, Canada.

^cRarely used in mastitis treatment and control.

^dThe concentration is in international units.

Agar disk diffusion method.

Bacteria were subcultured twice using a Columbia agar plate with 5% sheep blood (Oxoid Canada, Nepean, Ontario, Canada). Thereafter, the inocula were prepared for Sensititre and ADD tests. The ADD test was carried out on the basis of CLSI guidelines. In short, the inoculum was prepared in sterile demineralized water to a 0.5 McFarland turbidity standard for estimating cell density. Seeding of the Mueller-Hinton (MH) agar (Oxoid Canada, Nepean, Ontario, Canada) plate was done with the broth suspension using a cotton swab. Antimicrobial disks were then placed on the agar plates. Plates were incubated overnight (18 to 24 h) at 37°C (3, 20). Zone of inhibition diameters were measured in millimeters.

Sensititre system MIC method.

Pure culture, grown overnight on a Columbia agar plate with 5% sheep blood, was used for making a bacterial suspension in demineralized water for the Sensititre method. This suspension was standardized to a 0.5 McFarland turbidity standard and confirmed using the Sensititre nephelometer. Subsequently, a 10-μl aliquot was transferred using a calibrated loop into a tube of Sensititre Mueller-Hinton broth (MHB) that was finally mixed on a Vortex mixer (Vortex Manufacturing Company) for approximately 10 s. A Sensititre single-use dose head was placed on the Mueller-Hinton broth tube, and the tube was then placed in the Sensititre AutoInoculater according to the manufacturer's specifications. The AutoInoculater delivered 50 μl into each well containing serial 2-fold dilutions of antimicrobials on the bovine mastitis plate. After inoculation, the panel was covered with an adhesive seal and incubated overnight. MICs of different antimicrobial-bacterium combinations were determined manually. Afterwards, the same person recorded the automatic readings by using the Sensititre Auto Reader so as to prevent bias.

Manual broth microdilution test method. (i) Culture and inoculum preparation.

A computer-driven method of drawing observations randomly without replacement was used for selecting 119 isolates. These randomly selected isolates were streaked onto a Columbia agar plate with 5% sheep blood. All isolates were incubated at 35°C without CO₂, except for streptococci, which were incubated in the presence of CO₂ to obtain sufficient growth. Well-isolated colonies of fresh isolates (18 to 24 h) of the same morphological type were transferred from the agar plate and diluted in 2 ml of physiological saline to attain a 0.5 McFarland turbidity standard.

(ii) Stock solution preparation.

Reference powders of ampicillin, cephalothin, erythromycin, oxacillin, penicillin, novobiocin, and tetracycline were obtained commercially from Sigma-Aldrich Canada Ltd. (Oakville, Ontario, Canada) to prepare antimicrobial stock solutions. Ceftiofur and pirlimycin powders were obtained directly from the manufacturer (Pfizer Animal Health, Kirkland, Québec, Canada). All reference powders were stored as recommended by the manufacturers. The stock solutions were prepared to contain drug concentrations at four times the final concentrations of the highest concentration on the MIC panel.

The stock solutions were sterilized by filtering through a membrane filter. All of the stock solutions were dispensed in tubes and stored at −20°C, except for tetracycline, which was stored at 4°C. The tubes were removed as needed and used on the same day. Any unused solution was discarded at the end of the day. Antimicrobial drug concentrations encompassed both the quality control range and the CLSI breakpoints.

(iii) Medium preparation.

MHB (CM0405; Oxoid) was made following the manufacturer's instructions and was supplemented with Ca²⁺ and Mg²⁺ (MHBCM) after being autoclaved and chilled to 4°C. For streptococcal isolates, 5% laked horse blood was added to the MH broth (LMHBCM).

(iv) Microdilution tray preparation and inoculation.

Using a single pipette, 0.1 ml of double-strength MHBCM (42 g/liter of distilled water) was added to the first row, followed by further additions of 0.1 ml of single-strength MHBCM (21 g/liter of distilled water) to the remaining wells in the microdilution tray. Thereafter, 0.1 ml of the antimicrobial stock solution was added to the first row and later transferred to the remaining wells for serial dilutions, so that the final volume stayed at 0.1 ml in the wells. After the last dilution, 0.1 ml was discarded.

After dilution of the standardized inoculum to a 1:100 ratio with single-strength MHBCM, the inoculum was dispensed in the wells within 15 min. Each plate included the CLSI positive reference control well, as well as a series of inoculum-free dilution wells serving as a negative control. Quality control organisms used for positive controls included S. aureus ATCC 29213, E. faecalis ATCC 29212, E. coli ATCC 25922, and S. pneumoniae ATCC 49619. The plates were sealed and incubated for 18 to 24 h at 35°C. The MICs were determined on the basis of the presence or absence of turbidity in the wells. Two trained laboratory personnel recorded MICs afterwards.

To ascertain the final inoculum density and purity, 0.1 ml of the standardized inoculum was dispensed in 10 ml saline solution, and later, 0.1 ml was streaked out on a Columbia agar plate with 5% sheep blood. After incubation for 18 to 24 h, the plates were checked for purity and the colonies were counted.

Statistical analyses.

The manual broth microdilution test method was used as the reference method. MIC values (as determined by the Sensititre automatic and manual reading methods and the reference test method) outside the antimicrobial dilution range were defined as off-scale MICs, while the MICs within the dilution range were defined as on-scale or finite MICs. The off-scale MIC pairs were assumed to be in agreement for the sake of statistical analysis. Using CLSI guidelines, the isolates were classified sensitive, intermediate, or resistant.

The receiver operating characteristic (ROC) analysis methodology was used to assess the diagnostic accuracy of the Sensititre and ADD methods with reference to the manual broth microdilution test method (11). The area under the ROC curve (AUC) was used as a summary statistic. Intermediate category isolates were merged with the resistant category isolates to determine the AUC statistic. On the basis of the AUC statistic, the diagnostic test can be either noninformative (AUC = 0.5), less accurate (0.5 < AUC ≤ 0.7), moderately accurate (0.7 < AUC ≤ 0.9), highly accurate (0.9 < AUC < 1), or perfect (AUC = 1) (26).

Quantitative agreement for MICs was measured in terms of absolute and essential agreements between different test methods (Sensititre automatic MIC method compared to reference test method; the Sensititre manual reading method compared to automatic reading method), and the test statistic was the intertest MIC ratio (R). Absolute agreement was defined as the percentage of intertest MIC pairs with R equal to 1. R equal to 2 indicated 1 dilution underestimation, whereas R equal to 0.5 indicated 1 dilution overestimation by the Sensititre method in comparison to the MIC dilution for the manual broth microdilution test method (±1 log₂). Since a single 2-fold dilution was the inherent variability of the MIC dilution systems, intertest MICs within this tolerance range were considered to be in an essential agreement. In other words, essential agreement was defined as the percentage of intertest MIC pairs with R values of 0.5, 1, and 2. Errors were defined as intertest MIC pairs with R values of <0.5 or >2.

Limits-of-agreement (LOA) analysis was used to assess agreement between Sensititre automatic MIC reading and manual MIC reading methods as well (4). Limits-of-agreement values precisely quantified the differences between test methods by comparing the differences in log MIC values determined from the respective test methods with the mean of the log MICs.

The proportion agreement analysis method was used to ascertain the categorical agreement (sensitive, intermediate, and resistant) of the Sensititre automatic MIC and ADD methods with the reference manual broth microdilution method. Categorical agreement was defined as percent accordance between qualitative AMR profile results obtained using the Sensititre automatic MIC or ADD method and those obtained by the reference method. Very major error, major error, and discrepancy percentages were calculated for different isolate-drug combinations. A very major error was defined as an error in an AMR profile result if an isolate was categorized resistant by the reference test method and sensitive by the Sensititre or ADD method (false sensitive). A major error was defined as an error in an AMR profile result if an isolate was categorized sensitive by the reference test method and resistant by the Sensititre or ADD method (false resistant). Discrepancy was defined as an error in an AMR profile result if an intermediate isolate was categorized sensitive or resistant and vice versa. Some of the antimicrobial-isolate combinations were not analyzed for categorical agreement due to clinical inappropriateness and/or universal resistance (oxacillin, penicillin, erythromycin, and pirlimycin with E. coli) and a lack of interpretive criteria (oxacillin-enterococci/streptococci, tetracycline-enterococci, pirlimycin/ceftiofur-enterococci).

Data analyses were performed using the Intercooled Stata for Macintosh, version 10.1, program (Stata Corporation, College Station, TX).

RESULTS

Sensititre automatic MIC reading method compared with manual broth microdilution test.

The AUC estimates ranged from 0.27 to 1.00 (Table 2). Sensititre exhibited the lowest predictive accuracy for the S. aureus-ceftiofur combination. The predictive accuracy was less than 0.5 for 10% of the isolate-antimicrobial combinations. Overall, Sensititre was noninformative (AUC = 0.5), less accurate (0.5 < AUC ≤ 0.7), moderately accurate (0.7 < AUC ≤ 0.9), highly accurate (0.9 < AUC < 1), and perfect (AUC = 1) in its predictive accuracy for 7, 0, 53, 21, and 9% of the isolate-antimicrobial combinations, respectively.

Table 2.

Diagnostic accuracy estimates of Sensititre automatic readings (off- and on-scale MICs) and ADD test method with reference to manual broth microdilution test method for clinical bovine mastitis pathogens

Isolate (no. of isolates)a	Antimicrobial	AUC (95% CIb)
		Sensititre	ADD
Escherichia coli (n = 25)	Ampicillin	0.71 (0.50–0.87)	0.84 (0.63–0.95)
	Cephalothin	0.80 (0.59–0.93)	0.64 (0.42–0.82)
	Tetracycline	0.72 (0.50–0.87)	0.92 (0.73–0.99)
Staphylococcus aureus (n = 24)	Ampicillin	0.87 (0.67–0.97)	0.85 (0.62–0.95)
	Oxacillin	0.50 (0.29–0.70)	0.00 (0.00–0.14)c
	Penicillin	0.88 (0.67–0.97)	0.83 (0.62–0.95)
	Ceftiofur	0.27 (0.10–0.48)	0.05 (0.00–0.21)
Streptococcus uberis (n = 20)	Ampicillin	0.95 (0.75–0.99)	0.87 (0.62–0.96)
	Tetracycline	0.88 (0.68–0.98)	0.94 (0.75–0.99)
	Pirlimycin	1.00 (0.83–1.00)c	1.00 (0.83–1.00)c
Streptococcus dysgalactiae (n = 24)	Tetracycline	0.90 (0.73–0.98)	0.95 (0.75–0.99)
	Pirlimycin	0.94 (0.77–0.99)	1.00 (0.85–1.00)c
Other streptococci (n = 11)	Ampicillin	0.44 (0.16–0.76)	0.50 (0.23–0.83)
	Tetracycline	1.00 (0.71–1.00)c	1.00 (0.71–1.00)c
Enterococci (n = 15)	Cephalothin	0.98 (0.78–1.00)	0.89 (0.59–0.98)
	Penicillin	0.93 (0.68–0.99)	1.00 (0.78–1.00)c
	Erythromycin	0.90 (0.68–0.99)	0.73 (0.44–0.92)

^aA total of 119 isolates were tested. Isolates were either all sensitive or all resistant by the manual broth microdilution test method, and hence, no estimates are available for some antimicrobial-isolate combinations.

^bValues in parentheses are the 95% binomial exact confidence interval (95% CI), unless indicated otherwise.

^cOne-sided, 97.5% confidence interval.

Absolute agreement (R = 1) between respective MIC values (off and on scale) was evident for 12 to 100% of various isolate-antimicrobial combinations, the lowest being for the E. coli-cephalothin combination (Table 3). Essential agreement between test methods was evident for 16 to 100% of the isolate-antimicrobial combinations, with the lowest being for the E. coli-tetracycline combination. Among the various isolate-antimicrobial combinations, the underestimation bias was evident for 0 to 84% of the isolate-antimicrobial combinations, with the highest being for the E. coli-tetracycline combination. Overestimation bias was evident for 0 to 15% of different isolate-antimicrobial combinations. Overall, Sensititre exhibited underestimation and overestimation biases for 11 and 3% of the isolate-antimicrobial combinations, respectively; absolute agreement and essential agreement were evident for 75 and 87% of the isolate-antimicrobial combinations, respectively.

Table 3.

Essential agreement between Sensititre MIC automatic readings (off- and on-scale MICs) and manual broth microdilution MIC test method for clinical bovine mastitis pathogens

Isolate (no. of isolates)a	Antimicrobial	No. (%) of isolates with the following:
		Rb < 0.5	R = 0.5	R = 1	R = 2	R > 2	EAc	Errorsd
Escherichia coli (n = 25)	Ampicillin		1 (4)	4 (16)	10 (40)	10 (40)	15 (60)	10 (40)
	Oxacillin			25 (100)			25 (100)
	Cephalothin			3 (12)	13 (52)	9 (36)	16 (64)	9 (36)
	Penicillin-novobiocin			10 (40)	14 (56)	1 (4)	24 (96)	1 (4)
	Erythromycin			25 (100)			25 (100)
	Tetracycline			4 (16)		21 (84)	4 (16)	21 (84)
	Penicillin			21 (84)		4 (16)	21 (84)	4 (16)
	Ceftiofur			22 (92)		2 (8)	22 (92)	2 (8)
	Pirlimycin			24 (100)			24 (100)
Staphylococcus aureus (n = 24)	Ampicillin	1 (4)	1 (4)	17 (71)		5 (21)	18 (75)	6 (25)
	Oxacillin			23 (96)		1 (4)	23 (96)	1 (4)
	Cephalothin			24 (100)			24 (100)
	Penicillin-novobiocin			24 (100)			24 (100)
	Erythromycin	1 (4)	1 (4)	22 (92)			23 (96)	1 (4)
	Tetracycline			24 (100)			24 (100)
	Penicillin	2 (8)		17 (71)		5 (21)	17 (71)	7 (29)
	Ceftiofur		5 (22)	15 (61)	3 (13)	1 (4)	23 (96)	1 (4)
	Pirlimycin			24 (100)			24 (100)
Streptococcus uberis (n = 20)	Ampicillin		2 (10)	17 (85)	1 (5)		20 (100)
	Oxacillin			17 (85)		3 (15)	17 (85)	3 (15)
	Cephalothin	1 (5)		18 (90)	1 (5)		19 (95)	1 (5)
	Penicillin-novobiocin			20 (100)			20 (100)
	Erythromycin			20 (100)			20 (100)
	Tetracycline	1 (5)		16 (80)		3 (15)	16 (80)	4 (20)
	Penicillin	3 (15)	2 (10)	15 (75)			17 (85)	3 (15)
	Ceftiofur			15 (75)		5 (25)	15 (75)	5 (25)
	Pirlimycin			16 (80)		4 (20)	16 (81)	4 (20)
Streptococcus dysgalactiae (n = 24)	Ampicillin	2 (8)		22 (92)			22 (92)	2 (8)
	Oxacillin	1 (4)		23 (96)			23 (96)	1 (4)
	Cephalothin	3 (12)		21 (88)			21 (88)	3 (12)
	Penicillin-novobiocin	1 (4)		23 (96)			23 (96)	1 (4)
	Erythromycin	3 (12)		21 (88)			21 (88)	3 (12)
	Tetracycline	1 (4)		8 (33)	6 (25)	9 (38)	14 (58)	10 (42)
	Penicillin	2 (8)		22 (92)			22 (92)	2 (8)
	Ceftiofur			24 (100)			24 (100)
	Pirlimycin	1 (4)		20 (84)		3 (12)	20 (84)	4 (16)
Other streptococci (n = 11)	Ampicillin	1 (9)		8 (73)		2 (18)	8 (73)	3 (27)
	Oxacillin			11 (100)			11 (100)
	Cephalothin		1 (9)	9 (82)		1 (9)	10 (91)	1 (9)
	Penicillin-novobiocin			11 (100)			11 (100)
	Erythromycin			11 (100)			11 (100)
	Tetracycline			10 (91)	1 (9)		11 (100)
	Penicillin		1 (9)	10 (91)			11 (100)
	Ceftiofur			10 (91)		1 (9)	10 (91)	1 (9)
	Pirlimycin			11 (100)			11 (100)
Enterococci (n = 15)	Ampicillin		1 (7)	7 (46)	3 (20)	4 (27)	11 (73)	4 (27)
	Oxacillin			2 (13)	8 (54)	5 (33)	10 (67)	5 (33)
	Cephalothin		1 (7)	6 (40)	7 (46)	1 (7)	14 (93)	1 (7)
	Penicillin-novobiocin		6 (40)	3 (20)	5 (33)	1 (7)	14 (93)	1 (7)
	Erythromycin			6 (40)	7 (46)	2 (13)	13 (87)	2 (13)
	Tetracycline		1 (7)	2 (13)	11 (73)	1 (7)	14 (93)	1 (7)
	Penicillin		4 (27)	6 (40)	2 (13)	3 (20)	12 (80)	3 (20)
	Ceftiofur			10 (66)	1 (7)	4 (27)	11 (73)	4 (27)
	Pirlimycin	1 (7)		7 (46)	1 (7)	6 (40)	8 (53)	7 (47)
Overall		25 (2)	27 (3)	806 (75)	94 (9)	117 (11)	927 (87)	142 (13)

^aA total of 119 isolates were tested.

^bR, manual broth microdilution MIC/Sensititre automatic MIC ratio; R = 1, absolute intertest MIC agreement (no difference between manual broth microdilution MIC and Sensititre automatic MIC methods); R = 0.5 to 2, acceptable inherent variability (tolerance) range in MIC dilution systems; R < 0.5 and R > 2, over- and underestimation by Sensititre in reference to manual broth microdilution test method results, respectively.

^cEA, essential agreement (percentage of manual broth microdilution MIC and Sensititre automatic MIC pairs within tolerance range).

^dErrors, percentage of manual broth microdilution MIC and Sensititre automatic MIC pairs out of tolerance range.

Overall essential agreement values across Gram-positive isolates ranged from 79 to 95%, and the lowest value was evident for enterococcal isolates (Table 4). The overall essential agreement value for all Gram-positive isolates was 89%. Among Gram-positive isolates, essential agreement across different antimicrobials was the highest among other streptococci (95%), followed by S. aureus (93%), S. uberis (89%), S. dysgalactiae (88%), and enterococci (79%). Escherichia coli-different antimicrobial combinations had a far lower overall essential agreement value (79%).

Table 4.

Lower essential agreement between Sensititre MIC automatic readings (off- and on-scale MICs) and manual broth microdilution MIC test method for Escherichia coli isolates in comparison to Gram-positive bovine clinical mastitis isolates^a

Antimicrobial	Essential agreement (%)
	Staphylococcus aureus	Streptococcus uberis	Streptococcus dysgalactiae	Other streptococci	Enterococci	All Streptococcib	Gram-positive isolatesc	Escherichia coli
Ampicillin	75	100	92	73	73	88	83	60
Oxacillin	96	85	96	100	67	94	89	100
Cephalothin	100	95	88	91	93	91	93	64
Penicillin-novobiocin	100	100	96	100	93	99	98	96
Erythromycin	96	100	88	100	87	96	94	100
Tetracycline	100	80	58	100	93	79	86	16
Penicillin	71	85	92	100	80	92	86	84
Ceftiofur	96	75	100	91	73	89	87	92
Pirlimycin	100	81	84	100	53	88	84	100
Overall	93	89	88	95	79	91	89	79

^aA total of 119 isolates were tested.

^bAll streptococcal isolates include S. uberis, S. dysgalactiae, and other streptococci.

^cAll Gram-positive isolates include S. aureus, S. uberis, S. dysgalactiae, other streptococci, and enterococci.

Overall categorical agreement was 91%, with rates of very major errors, major errors, and discrepancies being 3, 1, and 5%, respectively. Categorical agreement between test methods ranged from 32 to 100% for specific isolate-antimicrobial combinations (Table 5). Intertest categorical agreement was the lowest for the E. coli-cephalothin combination, with very major errors and discrepancies occurring 20 and 48% of the time, respectively. Furthermore, the E. coli-ampicillin and S. dysgalactiae-tetracycline combinations had low categorical agreement values (72 and 50%, respectively). Escherichia coli-tetra-cycline, S. uberis-tetracycline/pirlimycin, and S. dysgalactiae-tetracycline/pirlimycin combinations had notably higher values of very major errors (≥10%).

Table 5.

Categorical agreement between Sensititre automatic readings (off- and on-scale MICs) and ADD test method with reference to manual broth microdilution test method for clinical bovine mastitis pathogens

Isolates (no. of isolates)a	No. (%) of isolates with the indicated categorical agreementb
	Sensititre				Agar disk diffusion
	CA	VME	ME	D	CA	VME	ME	D
Escherichia coli (25)
Ampicillin	18 (72)	2 (8)		5 (20)	14 (56)			11 (44)
Cephalothin	8 (32)	5 (20)		12 (48)	14 (56)	2 (8)	1 (4)	8 (32)
Tetracycline	20 (80)	3 (12)		2 (8)	22 (88)	1 (4)		2 (8)
Ceftiofur	25 (100)				25 (100)
Staphylococcus aureus (24)
Ampicillin	22 (92)	2 (8)			22 (92)	1 (4)	1 (4)
Oxacillin	23 (96)	1 (4)			23 (96)	1 (4)
Cephalothin	24 (100)				24 (100)
Penicillin-novobiocin	24 (100)				24 (100)
Erythromycin	24 (100)				20 (83)			4 (17)
Tetracycline	24 (100)				24 (100)
Penicillin	21 (88)	1 (4)	2 (8)	1 (4)	23 (96)			1 (4)
Ceftiofur	23 (96)			1 (4)	23 (96)			1 (4)
Pirlimycin	24 (100)				24 (100)
Streptococcus uberis (20)
Ampicillin	19 (95)			1 (5)	18 (90)			2 (10)
Cephalothin	20 (100)				20 (100)
Penicillin-novobiocin	20 (100)				20 (100)
Erythromycin	20 (100)				20 (100)
Tetracycline	16 (80)	2 (10)	1 (5)	1 (5)	16 (80)		1 (5)	3 (15)
Penicillin	20 (100)				20 (100)
Ceftiofur	20 (100)				20 (100)
Pirlimycin	17 (87)	3 (13)			20 (100)
Streptococcus dysgalactiae (24)
Ampicillin	22 (92)		1 (4)	1 (4)	24 (100)
Cephalothin	23 (96)		1 (4)		24 (100)
Penicillin-novobiocin	23 (96)		1 (4)		24 (100)
Erythromycin	21 (88)		2 (8)	1 (4)	24 (100)
Tetracycline	12 (50)	3 (12)		9 (37)	13 (55)		3 (12)	8 (33)
Penicillin	24 (100)				24 (100)
Ceftiofur	24 (100)				24 (100)
Pirlimycin	20 (84)	3 (12)	1 (4)		24 (100)
Other streptococci (11)
Ampicillin	9 (82)			2 (18)	9 (82)			2 (18)
Cephalothin	11 (100)				11 (100)
Penicillin-novobiocin	11 (100)				11 (100)
Erythromycin	11 (100)				11 (100)
Tetracycline	10 (91)			1 (9)	10 (91)			1 (9)
Penicillin	11 (100)				11 (100)
Ceftiofur	11 (100)				11 (100)
Pirlimycin	11 (100)				11 (100)
Enterococci (15)
Ampicillin	15 (100)				14 (93)			1 (7)
Cephalothin	13 (86)	1 (7)		1 (7)	7 (47)	3 (20)		5 (33)
Erythromycin	13 (86)			2 (14)	11 (73)			4 (27)
Penicillin	14 (93)	1 (7)			13 (86)	2 (14)
Overall	740 (91)	27 (3)	9 (1)	40 (5)	747 (92)	11 (1)	6 (1)	52 (6)

^aA total of 119 isolates were tested.

^bCA, categorical agreement (accordance for sensitive and resistant isolates between Sensititre automatic readings and agar disk diffusion test method with reference to manual broth microdilution test method); VME, very major error (an isolate categorized resistant by the reference method but classified sensitive by the Sensititre automatic reading/agar disk diffusion test method [false sensitive]); ME, major error (an isolate categorized sensitive by reference method but classified resistant by Sensititre automatic reading/agar disk diffusion test method [false resistant]); D, discrepancy (an intermediate isolate classified sensitive or resistant and vice versa).

Sensititre manual MIC reading method compared with automated reading method.

The Sensititre manual MIC reading method exhibited absolute and essential agreement with the automated reading method for 94 and 97% of the isolate-antimicrobial combinations, respectively. Absolute and essential agreements between respective MIC values (off and on scale) were evident for 76 to 100% and 83 to 100% of the various isolate-antimicrobial combinations, respectively. Notably, ≥10% errors were evident for the S. uberis-pirlimycin (17%), S. aureus-penicillin/ampicillin (12 and 10%, respectively), and enterococcus-ceftiofur/pirlimycin (10% each) combinations.

The magnitude of the mean bias between the Sensititre automatic and manual MIC reading methods ranged from an underestimation of 15% for the S. aureus-penicillin combination to an overestimation of 40% for the enterococci-pirlimycin combination (Table 6). Limits of agreement varied from an overestimation of 127% to an underestimation of 73% for the former combination and from an overestimation of 270% to an underestimation of 46% for the latter.

Table 6.

Differences between Sensititre automatic MIC readings (off- and on-scale MICs) and Sensititre manual MIC reading test method for clinical bovine mastitis pathogens

Isolate (no. of isolates)a	Antimicrobial	Mean R (95% CI)b	% biasc	Limits of Rd	Limits of agreemente
Escherichia coli (153)	Ampicillin	1 (0.98–1.02)		0.76–1.30	31 (+)–24 (−)
	Oxacillin	1 (1–1)		1–1
	Cephalothin	0.96 (0.93–0.99)	4 (+)	0.66–1.40	51 (+)–28 (−)
	Penicillin–novobiocin	0.94 (0.92–0.97)	6 (+)	0.65–1.37	53 (+)–27 (−)
	Ceftiofur	0.99 (0.99–1.00)	1 (+)	0.89–1.10	12 (+)–10 (−)
	Pirlimycin	1 (1–1)		1–1
	Erythromycin	1 (1–1)		1–1
	Tetracycline	0.91 (0.86–0.96)	9 (+)	0.45–1.80	122 (+)–44 (−)
	Penicillin	0.98 (0.97–1.00)	2 (+)	0.81–1.19	23 (+)–16 (−)
Staphylococcus aureus (151)	Ampicillin	1.13 (1.06–1.22)	11 (−)	0.47–2.74	112 (+)–63 (−)
	Oxacillin	1 (1–1)		1–1
	Cephalothin	1 (1–1)		1–1
	Penicillin-novobiocin	1 (1–1)		1–1
	Ceftiofur	0.98 (0.96–1.00)	2 (+)	0.71–1.34	40 (+)–26 (−)
	Pirlimycin	0.99 (0.98–1.00)	1 (+)	0.85–1.16	17 (+)–14 (−)
	Erythromycin	1.04 (1.00–1.08)	3 (−)	0.63–1.71	58 (+)–41 (−)
	Tetracycline	1.04 (1.00–1.08)	4 (−)	0.63–1.71	58 (+)–41 (−)
	Penicillin	1.17 (1.08–1.27)	15 (−)	0.44–3.70	127 (+)–73 (−)
Streptococcus uberis (47)	Ampicillin	1.04 (0.98–1.11)	4 (−)	0.66–1.64	51 (+)–39 (−)
	Oxacillin	0.94 (0.86–1.02)	6 (+)	0.53–1.66	88 (+)–39 (−)
	Cephalothin	1.10 (0.99–1.22)	10 (−)	0.54–2.24	85 (+)–55 (−)
	Penicillin-novobiocin	1 (1–1)		1–1
	Ceftiofur	1 (1–1)		1–1
	Pirlimycin	0.74 (0.63–0.88)	35 (+)	0.23–2.33	334 (+)–58 (−)
	Erythromycin	1.05 (0.98–1.15)	4 (−)	0.60–1.85	66 (+)–45 (−)
	Tetracycline	0.98 (0.96–1.01)	2 (+)	0.81–1.19	23 (+)–16 (−)
	Penicillin	1.16 (1.02–1.31)	14 (−)	0.50–2.66	100 (+)–63 (−)
Streptococcus dysgalactiae (49)	Ampicillin	1.05 (0.98–1.15)	5 (−)	0.60–1.85	66 (+)–45 (−)
	Oxacillin	1 (1–1)		1–1
	Cephalothin	1.11 (1–1.24)	9 (−)	0.52–2.41	92 (+)–58 (−)
	Penicillin-novobiocin	1 (1–1)		1–1
	Ceftiofur	0.98 (0.92–1.02)	2 (+)	0.65–1.43	53 (+)–30 (−)
	Pirlimycin	0.89 (0.81–0.99)	12 (+)	0.44–1.82	127 (+)–46 (−)
	Erythromycin	1.05 (0.88–1.25)	4 (−)	0.30–3.70	233 (+)–72 (−)
	Tetracycline	0.98 (0.88–1.09)	2 (+)	0.45–2.13	122 (+)–54 (−)
	Penicillin	1.10 (0.98–1.24)	10 (−)	0.48–2.50	108 (+)–60 (−)
Other streptococci (18)	Ampicillin	1 (1–1)		1–1
	Oxacillin	1 (1–1)		1–1
	Cephalothin	1 (1–1)		1–1
	Penicillin-novobiocin	1 (1–1)		1–1
	Ceftiofur	0.96 (0.88–1.04)	4 (+)	0.69–1.33	44 (+)–25 (−)
	Pirlimycin	0.93 (0.78–1.08)	7 (+)	0.48–1.76	108 (+)–44 (−)
	Erythromycin	1 (1–1)		1–1
	Tetracycline	1 (1–1)		1–1
	Penicillin	1 (1–1)		1–1
Enterococci (29)	Ampicillin	0.86 (0.77–0.97)	16 (+)	0.49–1.52	104 (+)–34 (−)
	Oxacillin	0.76 (0.61–0.95)	31 (+)	0.24–2.39	316 (+)–58 (−)
	Cephalothin	0.93 (0.86–1.01)	7 (+)	0.60–1.41	66 (+)–29 (−)
	Penicillin-novobiocin	0.97 (0.92–1.02)	3 (+)	0.75–1.26	33 (+)–20 (−)
	Ceftiofur	0.86 (0.72–1.03)	16 (+)	0.33–2.20	203 (+)–55 (−)
	Pirlimycin	0.71 (0.60–0.86)	40 (+)	0.27–1.84	270 (+)–46 (−)
	Erythromycin	0.93 (0.83–1.03)	7 (+)	0.53–1.63	88 (+)–38 (−)
	Tetracycline	0.97 (0.92–1.02)	3 (+)	0.75–1.26	33 (+)–20 (−)
	Penicillin	0.91 (0.83–0.99)	9 (+)	0.55–1.47	81 (+)–32 (−)

^aA total of 457 isolates were tested.

^bMean R, mean of Sensititre automatic MIC/manual MIC ratios; CI, confidence interval.

^cAbsolute bias, average overestimation (+) or underestimation (−) percentage by Sensititre manual MIC reading method compared to results of Sensititre automatic MIC reading method.

^dLimits of R, range of Sensititre automatic MIC/manual MIC ratios.

^eLimits of overestimation (+) and/or underestimation (−) percentage by Sensititre manual MIC test method compared to results of Sensititre automatic MIC reading method.

Agar disk diffusion test method.

The AUC estimates ranged from 0 to 1.00 (Table 2). Predictive accuracy of less than 0.5 was evident for 14% of the isolate-antimicrobial combinations. Overall, the ADD test method was noninformative, less accurate, moderately accurate, highly accurate, and perfect in its predictive accuracy for 3, 7, 36, 20, and 20% of the isolate-antimicrobial combinations, respectively.

Overall categorical agreement was 92%, with rates of very major errors, major errors, and discrepancies being 1, 1, and 6%, respectively. Very major errors, major errors, and discrepancies ranged from 0 to 20%, 0 to 12%, and 0 to 44%, respectively. The enterococcus-cephalothin and E. coli-ampicillin combinations had the highest percentages of very major errors and discrepancies, respectively (20 and 44%). Notably, lower categorical agreement values were evident for the E. coli-cephalothin/ampicillin (56% each), S. dysgalactiae-tetracycline (55%), and enterococcus-cephalothin combinations (47%) (Table 5).

DISCUSSION

The primary objective of this study was to assess the diagnostic accuracy of the Sensititre MIC mastitis panel (Sensititre) and the ADD method of Bauer et al. (3) with reference to the manual broth microdilution test method for antimicrobial resistance profiling of udder pathogens. The study was designed to account for potential variation in susceptibility prevalence due to geographical and epidemiological differences. To the best of the knowledge of the authors, the Sensititre and the ADD methods have not been compared with a reference manual broth microdilution test method for multiple species of the most common clinical bovine mastitis pathogens.

Quantitative methods of antimicrobial susceptibility testing include agar dilution, broth macrodilution, and broth microdilution. Of these, standardized agar dilution test is traditionally considered the “gold standard” for antimicrobial susceptibility testing (28). However, due to its cumbersome nature and shorter shelf life of the agar (25) and broth macrodilution testing methods, broth microdilution is commonly used as a reference method for antimicrobial susceptibility testing. It is an efficient method, and decreased volumes of antimicrobials are used to attain results equivalent to those obtained with the standardized macrodilution method (8).

In the present study, intertest off-scale MIC pairs were assumed to be in essential agreement. However, the study results would not be valid if the data were not analyzed for finite-scale MICs. Therefore, to avoid bias in the study, data were analyzed for finite-scale as well for off- and on-scale MICs. Further, the data analysis was performed for individual isolate-drug combinations; Hansen and Freedy (12) made a similar recommendation in their study as well.

The area under the ROC curve was used as a summary estimate of the diagnostic accuracy of the Sensititre and the ADD methods in reference to the manual broth microdilution test method (11). This estimate indicates the ability of a diagnostic test (Sensititre/ADD) to discriminate between sensitive and resistant isolates (as determined by the reference method) over a range of values of a discriminatory variable (MIC/zone diameter). The AUC estimate can be interpreted as the probability of a higher MIC/lower zone diameter value for a randomly chosen resistant isolate over the value for a sensitive isolate. AUC estimates could not be determined for some isolate-antimicrobial combinations in the present study, as the isolates were either all sensitive or all resistant as determined by the reference method.

The limits-of-agreement analysis method was employed to quantify precisely the differences in MICs obtained using the Sensititre automatic and manual reading methods. This method provides a finer approach to compare quantitative agreement between test methods by determining the magnitude of intertest bias (1 minus the mean R value) and thereafter the limits up to which 95% of the intertest differences could vary. In certain cases of specific antimicrobial-isolate combinations where the intertest MIC differences exceeded the acceptable inherent variability range of the MIC dilution systems, the two methods should not be used interchangeably. Use of this methodology is more appropriate than use of the product-moment correlation coefficient (r) method because r measures the strength of an association between two methods and not the agreement (4). Subsequently, if the LOA values for different isolate-antimicrobial MIC combinations are within the tolerance range of a single 2-fold dilution, Sensititre automatic and manual reading methods can be used interchangeably.

Sensititre automatic MIC reading method.

Overall, Sensititre demonstrated a moderate to high predictive accuracy for the majority of the different isolate-antimicrobial combinations compared to the results of the manual broth microdilution MIC method. There were no AUC estimates for other mastitis pathogen-antimicrobial combinations from previous studies available for comparison.

Sensititre demonstrated very high absolute and essential agreements with the reference method for off- and on-scale MICs. However, Sensititre exhibited a profound inclination toward underestimation across different isolate-antimicrobial combinations for off- and on-scale MICs and finite MICs as well. Even within the tolerance range of MICs, Sensititre demonstrated an increased inclination toward underestimation. Essential agreement between test methods for Gram-positive isolate–beta-lactam combinations was similar to that of Gavan et al. (9) (86.7% and 87.8%, respectively). However, Sensititre demonstrated underestimation for 10% of these isolate-antimicrobial combinations in the present study, unlike the overestimation in the latter study. The lowest essential agreement percentage was evident for the E. coli-tetracycline combination in the present study (16%); it was far lower than the lowest essential agreement of 86.7% for the staphylococcus-penicillin G combination in the study by Gavan et al. (9).

When essential agreements between all Gram-positive and E. coli isolates are compared, Sensititre demonstrated an overall lower essential agreement for the latter. Sensititre demonstrated consistently lower essential agreement results for E. coli isolates than for all Gram-positive isolates, especially for ampicillin and cephalothin. Notably, the E. coli-tetracycline combination had the lowest essential agreement (16%). Furthermore, Sensititre exhibited a higher underestimation proportion for E. coli-antimicrobial combinations than for all Gram-positive isolates (27% and 9%, respectively); on the contrary, no overestimation was evident for the former. Therefore, this bias of Sensititre with E. coli isolates should be kept in mind while performing MIC testing. Even within the all Gram-positive isolate category, Sensititre demonstrated higher essential agreement for S. aureus and all streptococcal isolates than for enterococci. Notably, the lowest essential agreement within Gram-positive isolates was evident for the enterococcus-pirlimycin combination.

Even though the categorical agreement between test methods was very high for most pathogen-antimicrobial combinations, some of the isolate-antimicrobial combinations exhibited a very high percentage of very major errors, notably, E. coli-cephalothin and E. coli-tetracycline (20% and 12%, respectively), S. uberis-pirlimycin (13%), and S. dysgalactiae-tetracycline and -pirlimycin (12% each). Gradus had demonstrated that the percentages of very major errors and major errors should be less than 1% and 5%, respectively (10). Thus, susceptibility results from Sensititre for these pathogen-antimicrobial combinations should be interpreted with caution. Furthermore, Sensititre exhibited a higher percentage of categorical agreement results for ampicillin and cephalothin for all Gram-positive isolates than for E. coli, again demonstrating a bias against E. coli microorganisms for such antimicrobial-isolate combinations.

Sensititre automatic MIC reading test method in comparison to manual MIC reading test method.

Except for sulfadimethoxine, overall absolute and essential agreements between test methods for off- and on-scale MICs were very high for different isolate-antimicrobial combinations (94% and 97%, respectively). The Sensititre manufacturer's instructions for interpretation of the sulfadimethoxine MIC state that they must be manually read (27); thus, automatic MIC readings for this antimicrobial are not valid. The lowest absolute agreement was evident for enterococcus-ceftiofur and enterococcus-pirlimycin combinations. The magnitude of intertest MIC bias was measured, and the limits of agreement up to which 95% of intertest MIC differences (or R, Sensititre automatic MIC/manual MIC) could vary were determined as well. The acceptable limits of agreement between test methods are single 2-fold dilutions (±1 log₂ dilution). In the present study, the wider limits-of-agreement values exceeding the tolerance range of MICs between test methods for the E. coli-tetracycline, S. aureus-ampicillin/penicillin, S. uberis-pirlimycin/penicillin, S. dysgalactiae-pirlimycin/erythromycin/tetracycline, and enterococcus-ampicillin/oxacillin/ceftiofur/pirlimycin combinations indicate the need for caution while using manual MIC readings. Therefore, manual MIC readings should be avoided, when possible, for these isolate-antimicrobial combinations.

Agar disk diffusion method compared to manual broth microdilution test method.

Except for a few isolate-antimicrobial combinations, the ADD method demonstrated moderate to perfect predictive accuracy across the majority of isolate-antimicrobial combinations. However, the ADD test method was noninformative for S. aureus-oxacillin/ceftiofur and other streptococcus-ampicillin combinations. There were no AUC estimates for other mastitis pathogen-antimicrobial combinations from previous studies available for comparison. Further, the ADD test method demonstrated a lower categorical percentage for the E. coli-ampicillin and E. coli-cephalothin combinations than for combinations with Gram-positive isolates, a finding similar to that for Sensititre, thereby indicating a biased approach toward these isolate-antimicrobial combinations. Interestingly, the very major error percentage was the highest for enterococcus-various antimicrobials combinations; the slow growth rate of enterococci in MH agar medium in the ADD test method could be a potential reason for this observation (15). Relatively higher proportions of errors were encountered for various isolate-tetracycline combinations; variations in divalent cations such as calcium and magnesium in the MH agar medium could be a potential reason (5).

The overall categorical agreement percentage for the S. aureus-ampicillin/penicillin combinations was higher in the present study than in that of Schlegelova et al. (25); categorical agreement results for the S. aureus-cephalothin combination were similar. A relatively higher percent agreement for the sensitive S. aureus-ampicillin/penicillin combinations than for resistant ones was evident in the present study, a finding contrary to that of Schlegelova et al. (25). The percentage of very major errors for these isolate-antimicrobial combinations was lower in the present study than in the latter one.

Conclusions.

Sensititre demonstrated a range of predictive accuracy of from 71 to 99% for 74% of various isolate-antimicrobial combinations; the agar disk diffusion method demonstrated a range of predictive accuracy of from 71 to 100% for 76% of the isolate-antimicrobial combinations. However, both of these diagnostic tests demonstrated bias against E. coli isolates in comparison to the Gram-positive isolates, notably, for ampicillin and cephalothin antimicrobials. Even among Gram-positive isolates, Sensititre demonstrated higher essential agreement for S. aureus and all streptococcal isolates than for enterococci. Caution should therefore be employed while interpreting antimicrobial susceptibility test results in such cases. While the results are similar for most antimicrobial-isolate combinations, the Sensititre automatic reading method is more accurate for most specific isolate-antimicrobial combinations. Overall, both Sensititre and the agar disk diffusion test method demonstrated higher diagnostic agreement than diagnostic accuracy for the majority of isolate-antimicrobial combinations.