Evaluation of Antimicrobial Susceptibility Testing Methods for Burkholderia cenocepacia and Burkholderia multivorans Isolates from Cystic Fibrosis Patients

ABSTRACT

The Burkholderia cepacia complex (BCC) is known for causing serious lung infections in people with cystic fibrosis (CF). These infections can require lung transplantation, eligibility for which may be guided by antimicrobial susceptibility testing (AST). While the Clinical and Laboratory Standards Institute recommends AST for BCC, the European Committee on Antimicrobial Susceptibility Testing (EUCAST) does not, due to poor method performance and correlation with clinical outcomes. Furthermore, limited data exist on the performance of automated AST methods for BCC. To address these issues, reproducibility and accuracy were evaluated for disk diffusion (DD), broth microdilution (BMD), and MicroScan WalkAway using 50 B. cenocepacia and 50 B. multivorans isolates collected from people with CF. The following drugs were evaluated in triplicate: chloramphenicol (CAM), ceftazidime (CAZ), meropenem (MEM), trimethoprim-sulfamethoxazole (TMP-SMX), minocycline (MIN), levofloxacin (LVX), ciprofloxacin (CIP), and piperacillin-tazobactam (PIP-TAZ). BMD reproducibility was ≥ 95% for MEM and MIN only, and MicroScan WalkAway reproducibility was similar to BMD. DD reproducibility was < 90% for all drugs tested when a 3 mm cut-off was applied. When comparing the accuracy of DD to BMD, only MEM met all acceptance criteria. TMP-SMX and LVX had high minor errors, CAZ had unacceptable very major errors (VME), and MIN, PIP-TAZ, and CIP had both unacceptable minor errors and VMEs. For MicroScan WalkAway, no drugs met acceptance criteria. Analyses also showed that errors were not attributed to one species. In general, our data agree with EUCAST recommendations.

INTRODUCTION

The Burkholderia cepacia complex (BCC) is a group of 24 closely related Gram-negative bacteria that cause pulmonary infections in adults with cystic fibrosis (CF) lung disease. Approximately 3% of individuals with CF have BCC infections (1), and the median age at first BCC infection is 19.9 years old (2). BCC infections are most commonly caused by B. cepacia, B. cenocepacia, and B. multivorans, while another non-BCC Burkholderia species, B. gladioli, is also quite common; the remaining 21 species in the Burkholderia cepacia complex are relatively rarely isolated from people with CF (2). People with advanced CF lung disease often require lung transplantation due to acute clinical decline (3). While less common than other respiratory pathogens, BCC infections are especially problematic because they cause more rapid pulmonary function decline than Pseudomonas aeruginosa infections (4), and as a result, BCC pulmonary infections are more prevalent in people with CF requiring lung transplant (1, 2, 5). Additionally, infections by some BCC species have poor prognosis and lead to worse patient survival following lung transplant. For example, people infected with either non-epidemic strains of B. cenocepacia or the non-BCC species B. gladioli have reduced post-transplant survival compared to those infected with B. multivorans; therefore, B. cenocepacia or B. gladioli infection is often considered a contraindication for lung transplant (6). Regardless of the infecting species, infection management post-transplant is key to long-term patient survival. While eligibility criteria for lung transplant varies by clinic, most institutions will require at least one active drug.

Treatment of BCC infections is challenging due to intrinsic and acquired resistance (7–14), the limited number of drugs with established breakpoints (15), and different recommendations for antimicrobial susceptibility testing (AST) by standards-setting bodies. Indeed, a major AST survey found that the recommended antibiotics for BCC treatment were only effective against 3–38% of isolates tested from CF patients (16). The Clinical and Laboratory Standards Institute (CLSI) recommends AST for BCC and established broth microdilution (BMD) MIC breakpoints for ceftazidime (CAZ), chloramphenicol (CAM), levofloxacin (LVX), meropenem (MEM), minocycline (MIN), trimethoprim-sulfamethoxazole (TMP-SMX), and ticarcillin-clavulanate and disk diffusion (DD) breakpoints for CAZ, MEM, MIN, and TMP-SMX (15). In contrast, the European Committee on Antimicrobial Susceptibility Testing (EUCAST) does not recommend AST to guide therapy due to poor correlation between MIC and clinical outcomes and inconsistent susceptibility results among different AST methods (17). These findings are troubling because clinicians may use AST to determine whether BCC infections can be controlled pre- and post-lung transplant (6, 18), and patients with organisms that test resistant to all available drugs may encounter problems with transplant eligibility.

The general utility of AST in CF treatment has been questioned in recent years. The Antimicrobial Resistance (AMR) in CF International Working Group has recommended deprioritizing routine AST for managing pulmonary infections in individuals with CF (19). This recommendation was based on findings that, in general, AST results do not correlate with CF treatment outcomes for Pseudomonas aeruginosa lung infections (19). However, it remains that many clinicians will choose to use AST to guide their treatment regimens. Furthermore, in certain cases, such as for those patients with bloodstream infections, AST will remain an essential part of patient care. It is therefore important to ensure that AST provides clinicians with reproducible and accurate data.

Two previous studies have examined the accuracy and reproducibility of AST methods for BCC. Wootton et al. performed DD, BMD, agar dilution, and gradient diffusion on 155 BCC isolates from Europe. All isolates were from individuals with CF and were a mix of BCC species, the majority being B. cenocepacia and B. cepacia (20). The authors found poor reproducibility and accuracy for DD, BMD, agar dilution, and gradient diffusion methods when testing CAM, CAZ, MEM, MIN, TMP-SMX, and ciprofloxacin (CIP), with no drugs meeting all acceptance criteria (20). Fehlberg et al. evaluated BMD, DD, agar dilution, and gradient diffusion for CAM, CAZ, MEM, MIN, LVX, and TMP-SMX against 82 BCC isolates collected from multiple sources at two Brazilian hospitals (21). Forty-seven isolates were from individuals with CF, while 35 isolates were from non-CF patients. The authors found slightly better performance when comparing DD and BMD; MEM and MIN met all acceptance criteria, but CAZ, LVX, and TMP-SMX did not (21). The findings from these studies demonstrate the need for evaluation of AST methods for isolates from the United States (U.S.). Moreover, these studies analyzed smaller numbers of B. multivorans isolates, which is one of the three most common BCC species infecting people with CF in the United States (2) and a species that is not contraindicated for lung transplant.

Because CLSI recommends AST for BCC, the goal of this study was to evaluate AST methods for BCC isolates from people with CF in the United States. We performed BMD and DD assays in triplicate following current CLSI guidelines on 100 BCC isolates. In contrast to previous studies, all isolates were from individuals with CF in the United States. We focused our studies on B. cenocepacia (n = 50) and B. multivorans (n = 50), two of the three most frequently isolated BCC species in the U.S (2). When combined, these species can account for up to 69% of Burkholderia species infections (2, 6). Furthermore, lung transplant outcomes are most well understood for these species, where B. multivorans and B. cenocepacia are associated with better and worse survival post lung transplant, respectively (6, 22). Additionally, previous studies evaluating automated AST methods tested only small numbers of BCC isolates (23–25). Since these methods are used by some laboratories for BCC AST, we performed MicroScan WalkAway on the same BCC isolates in triplicate.

MATERIALS AND METHODS

Isolates.

One hundred BCC isolates were obtained from the Cystic Fibrosis Foundation Burkholderia cepacia Research Laboratory and Repository at the University of Michigan. Isolates were identified to the species level using recA-based PCR and recA sequence analysis as described previously (26, 27). The isolates included 50 B. cenocepacia and 50 B. multivorans isolates. All isolates were collected from patients in the United States. Isolates were selected from unique individuals with CF to reduce the chance of selecting clonal organisms. To represent contemporary isolates and those commonly observed in clinical practice, the most recent isolate recovered from each chronically infected person was selected.

DD and BMD.

DD and BMD were performed at the Clinical Microbiology Institute (Wilsonville, OR) following CLSI methods. Isolates were stored at −70°C, subcultured from frozen stock (F1 isolate) onto 5% sheep blood agar (SBAP; Thermofisher/Remel, Lexana, KS), and incubated at 35°C in ambient air for 20–24 h or until visible colonies were observed. F1 isolates were subcultured (F2 isolate) onto 5% SBAP for DD and BMD studies and incubated at 35°C in ambient air for 20–24 h. Replicates were performed on consecutive days, with each replicate having its own F1 and F2 subcultures. The following drugs with BCC breakpoints were tested by BMD: CAM, CAZ, LVX, MEM, MIN, and TMP-SMX. Limited data for intrinsic resistance to piperacillin-tazobactam (PIP-TAZ) exists (15), and in some cases CIP is used in combination therapy to treat BCC infections (28); therefore, CIP and PIP-TAZ were also tested, and P. aeruginosa breakpoints were applied. For DD, all the above drugs except CAM were tested. BMD was performed using custom panels made by the Clinical Microbiology Institute. Cation-adjusted Mueller-Hinton II broth powder was obtained from Becton, Dickinson (Sparks, MD). Antibiotics were tested in 2-fold serial dilutions with the following ranges: 0.06 μg/ml to 128 μg/ml, CAZ; 0.03 μg/ml to 64 μg/ml, MEM, MIN, LVX, CAM, CIP; 0.008/0.14 μg/ml to 16/304 μg/ml, TMP-SMX; and 0.25/4 μg/ml to 128/4 μg/ml, PIP-TAZ. Mueller-Hinton agar and disks for DD were obtained from Thermofisher/Remel. Growth media and antibiotics for all DD and BMD tests were from single manufacturer lots. Isolated colonies from F2 subcultures grown for 20–24 h on SBAP at 35°C in ambient air were suspended in saline (Thermofisher/Remel) according to CLSI guidelines to obtain a suspension with turbidity equivalent to a 0.5 McFarland standard (15). On each of the three testing days, the same standardized bacterial suspension was used for the DD and BMD inoculum for each isolate. BMD and DD plates were inoculated according to CLSI guidelines and incubated at 35°C in ambient air for 20–24 h. Plates were routinely read at 20 h and re-incubated for another 4 h; the final reading was 24 h. Zones of inhibition and MICs were read by two separate operators. On reading days, either one operator read the BMD MIC while the other measured the DD zone of inhibition, or a single operator read both DD and BMD results. Escherichia coli ATCC 25922, E. coli ATCC 35218, and P. aeruginosa ATCC 27853 were used for quality control, which was performed each day of testing.

It was noted during these studies that 21 isolates grew more slowly compared to the other 79 isolates. These isolates demonstrated lighter growth in DD and BMD assays. While slower growing, definite growth was always observed in the BMD control well, and a confluent lawn was visible on DD assays. Because slow-growing CF isolates are frequently encountered, and these isolates met CLSI criteria for interpreting zones of inhibition and MIC, we included them in our final analyses.

MicroScan WalkAway.

MicroScan WalkAway (Beckman Coulter, Inc., Brea, CA) was performed in the Department of Pathology and Clinical Microbiology at Duke University School of Medicine. Per the manufacturer, MicroScan WalkAway is FDA-cleared for reporting BCC susceptibility to CAZ, MEM, LVX, and TMP-SMX (personal communication). CIP and PIP-TAZ were tested off-label. MicroScan WalkAway assays were performed separately from the above BMD and DD assays. F1 and F2 isolates were prepared for testing as described above (15). MicroScan panels were prepared following the manufacturer’s instructions. Panels were inoculated using MicroScan water and a turbidity meter to make bacterial suspensions equivalent to a 0.5 McFarland Standard. Isolates were inoculated into respective MicroScan NM45 MIC panels and incubated and read by the MicroScan WalkAway system at 18 h. For each replicate, panels were from a different manufacturer lot. The results from all MIC trays were manually confirmed. Isolates were tested in triplicate, with each replicate being tested by a different technologist on different days.

Data analysis.

Zone diameters and MIC values were interpreted using the CLSI M100 Performance Standards for Antimicrobial Susceptibility Testing, 30th Edition, 2020 (Table 1) (15). P. aeruginosa breakpoints were applied when no BCC breakpoints were established, which included MIC breakpoints for CIP and PIP-TAZ (BMD and MicroScan WalkAway) and zone diameter breakpoints for LVX, CIP, and PIP-TAZ (DD).

TABLE 1.

CLSI breakpoints applied in this study

Organism(s)	Antimicrobial agent	Interpretive categories and zone diam breakpoints (mm)			Interpretive categories and MIC breakpoints (μg/ml)
		S	I	R	S	I	R
Burkholderia cepacia complex	CAM	–	–	–	≤8	16	≥32
	CAZ	≥21	18–20	≤17	≤8	16	≥32
	MEM	≥20	16–19	≤15	≤4	8	≥16
	TMP-SMX	≥16	11–15	≤10	≤2/38	–	≥4/76
	LVX	–	–	–	≤2	4	≥8
	MIN	≥19	15–18	≤14	≤4	8	≥16
Pseudomonas aeruginosa	LVX	≥22	15–21	≤14	–	–	–
	CIP	≥25	19–24	≤18	≤0.5	1	≥2
	PIP-TAZ	≥21	15-20	≤14	≤16/4	32/4-64/4	≥128/4

BMD reproducibility was defined as the number of isolates that had MICs the same as or within ± 1 log₂ dilution between replicates. The acceptability cut-off for BMD reproducibility was ≥ 95%, which was based on a previous study of BMD reproducibility for BCC (20). Only isolates with on-scale MICs were included in the analysis. For DD reproducibility, zone of inhibition ranges were calculated by comparing the measurements from the three replicates for each isolate. Ranges within 3 mm or 6 mm were recorded, as has previously been reported for Pseudomonas aeruginosa (29). MicroScan WalkAway reproducibility was analyzed as described for BMD; MIN was not evaluated because it was not included on the MicroScan WalkAway panel, and TMP-SMX could not be analyzed because only two values are reported, which are at the limit of detection for the assay.

To determine the accuracy of DD and MicroScan WalkAway, these methods were compared to the reference standard, BMD. For DD, CAM was not evaluated because DD breakpoints for neither BCC nor P. aeruginosa are established. CLSI recommends using the less stringent error-rate bounded method to analyze discrepancy rates for a given antibiotic if > 20% of the isolates have MIC values within ± 1 log₂ dilution of the established breakpoints (30, 31). The error-rate bounded method separates isolates into 3 groups based on their MICs to determine types of error: isolates with MICs ≥ 2 log₂ (≥I + 2) dilutions above the intermediate breakpoint, isolates with MICs ± 1 log₂ (I + 1 to I − 1) dilution of the intermediate breakpoint, and isolates with MICs ≤ 2 log₂ (≤I – 2) dilutions below the intermediate breakpoint. For all seven antibiotics in this comparison, between 28 and 89% of the 100 isolates had MICs within ± 1 log₂ dilution of the established breakpoints. Therefore, the error-rate bounded method was used as previously described to calculate very major errors (VME), major errors (ME), and minor errors (mE) (31). Acceptance criteria for the error-rate bounded method are shown in Tables 4–7. The mode MIC and mode zone of inhibition were used for calculating VMEs, MEs, and mEs. If no mode MIC or zone of inhibition existed, the median MIC or zone of inhibition was used. VMEs occurred when an isolate tested resistant by BMD but susceptible by DD or MicroScan WalkAway. MEs occurred when an isolate tested susceptible by BMD but resistant by DD or MicroScan WalkAway. mEs occurred when an isolate tested susceptible or resistant by BMD and intermediate by DD or MicroScan WalkAway or vice versa. Because errors were based on comparison of mode or median values from three replicate experiments, errors were not investigated further.

TABLE 4.

Comparison of DD to BMD using the error-rate bounded method

Antibiotic	Type	mE	ME	VME
CAZ	≥I +2	1/37 (2.7%)	NAa	3/37 (8.1%) b
	I + 1 TO I-1	11/38 (28.9%)	0/38 (0.0%)	3/38 (7.9%)
	≤I-2	0/25 (0.0%)	0/25 (0.0%)	NA
MEM	≥I +2	0/34 (0.0%)	NA	0/34 (0.0%)
	I + 1 TO I-1	22/61 (36.1%)	0/61 (0.0%)	3/61 (4.9%)
	≤I-2	0/5 (0.0%)	0/5 (0.0%)	NA
TMP-SMX	≥R +1	0/44 (0.0%)	NA	0/44 (0.0%)
	R + S	6/26 (23.1%)	1/26 (3.8%)	0/26 (0.0%)
	≤S -1	2/30 (6.7%)	0/30 (0.0%)	NA
MIN	≥I +2	11/33 (33.3%)	NA	5/33 (15.2%)
	I + 1 TO I-1	30/63 (47.6%)	0/63 (0.0%)	13/63 (20.6%)
	≤I-2	0/4 (0.0%)	0/4 (0.0%)	NA
LVXc	≥I +2	1/57 (1.8%)	NA	0/57 (0.0%)
	I + 1 TO I-1	15/39 (38.5%)	0/39 (0.0%)	0/39 (0.0%)
	≤I-2	1/4 (25.0%)	0/4 (0.0%)	NA
CIPd	≥I +2	8/89 (9.0%)	NA	0/89 (0.0%)
	I + 1 TO I-1	5/11 (45.5%)	0/11 (0.0%)	2/11 (18.2%)
	≤I-2	0/0 0.0%)	0/0 (0.0%)	NA
PIP-TAZd	≥IHi +2	7/49 (14.3%)	NA	5/49 (10.2%)
	IHi+1 TO ILo-1	23/32 (71.9%)	0/32 (0.0%)	5/32 (15.6%)
	≤ILo-2	0/19 (0.0%)	0/19 (0.0%)	NA
Acceptable criteria	≥I +2	<5%	NA	<2%
	I + 1 TO I-1	<40%	<10%	<10%
	≤I-2	<5%	<2%	NA

^aNA, not applicable.

^bBoldface indicates when acceptance criteria were not met.

^cDD zone diameters were interpreted using P. aeruginosa breakpoints.

^dBMD MICs and DD zone diameters were interpreted using P. aeruginosa breakpoints.

TABLE 5.

Comparison of MicroScan WalkAway to BMD using the error-rate bounded method

Antibiotic	Type	mE	ME	VME
CAZ	≥I +2	2/37 (5.4%) a	NAb	8/37 (21.6%)
	I + 1 TO I-1	6/38 (15.8%)	2/38 (5.3%)	8/38 (21.1%)
	≤I-2	0/25 (0.0%)	6/25 (24.0%)	NA
MEM	≥I +2	8/34 (23.5%)	NA	21/34 (61.8%)
	I + 1 TO I-1	22/61 (36.1%)	0/61 (0.0%)	30/61 (49.2%)
	≤I-2	0/5 (0.0%)	0/5 (0.0%)	NA
TMP-SMX	≥R +1	NA	NA	13/44 (29.5%)
	R + S	NA	3/26 (11.5%)	12/26 (46.2%)
	≤S -1	NA	5/30 (16.7%)	NA
LVX	≥I +2	6/57 (10.5%)	NA	17/57 (29.8%)
	I + 1 TO I-1	22/39 (56.4%)	0/39 (0.0%)	7/39 (17.9%)
	≤I-2	0/4 (0.0%)	2/4 (50.0%)	NA
PIP-TAZc	≥IHi +2	15/49 (30.6%)	NA	18/49 (36.7%)
	IHi+1 TO ILo-1	20/32 (62.5%)	0/32 (0.0%)	8/32 (25.0%)
	≤ILo-2	0/19 (0.0%)	3/19 (15.8%)	NA
Acceptable criteria	≥I +2	<5%	NA	<2%
	I + 1 TO I-1	<40%	<10%	<10%
	≤I-2	<5%	<2%	NA

^aBoldface indicates when acceptance criteria were not met.

^bNA, not applicable.

^cBMD and MicroScan WalkAway MICs were interpreted using P. aeruginosa breakpoints.

TABLE 6.

Comparison of DD to BMD using the error-rate bounded method by species

		B. cenocepacia			B. multivorans
Antibiotic	Type	mE	ME	VME	mE	ME	VME
CAZ	≥I +2	0/17 (0.0%)	NAa	3/17 (17.6%) b	1/20 (5.0%)	NA	0/20 (0.0%)
	I + 1 TO I-1	5/22 (22.7%)	0/22 (0.0%)	1/22 (4.5%)	6/16 (37.5%)	0/16 (0.0%)	2/16 (12.5%)
	≤I-2	0/11 (0.0%)	0/11 (0.0%)	NA	0/14 (0.0%)	0/14 (0.0%)	NA
MEM	≥I +2	0/11 (0.0%)	NA	0/11 (0.0%)	0/23 (0.0%)	NA	0/23 (0.0%)
	I + 1 TO I-1	9/35 (25.7%)	0/35 (0.0%)	1/35 (2.9%)	13/26 (50.0%)	0/26 (0.0%)	2/26 (7.7%)
	≤I-2	0/4 (0.0%)	0/4 (0.0%)	NA	0/1 (0.0%)	0/1 (0.0%)	NA
TMP-SMX	≥R +1	0/22 (0.0%)	NA	0/22 (0.0%)	0/22 (0.0%)	NA	0/22 (0.0%)
	R + S	3/14 (21.4%)	0/14 (0.0%)	0/14 (0.0%)	3/12 (25.0%)	1/12 (8.3%)	0/12 (0.0%)
	≤S -1	1/14 (7.1%)	0/14 (0.0%)	NA	1/16 (6.3%)	0/16 (0.0%)	NA
MIN	≥I +2	7/21 (33.3%)	NA	4/21 (19.0%)	4/12 (33.3%)	NA	1/12 (8.3%)
	I + 1 TO I-1	10/27 (37.0%)	0/27 (0.0%)	9/27 (33.3%)	20/36 (55.6%)	0/36 (0.0%)	4/36 (11.1%)
	≤I-2	0/2 (0.0%)	0/2 (0.0%)	NA	0/2 (0.0%)	0/2 (0.0%)	NA
LVXc	≥I +2	1/30 (3.3%)	NA	0/30 (0.0%)	0/27 (0.0%)	NA	0/27 (0.0%)
	I + 1 TO I-1	7/17 (41.2%)	0/17 (0.0%)	0/17 (0.0%)	8/22 (36.4%)	0/22 (0.0%)	0/22 (0.0%)
	≤I-2	0/3 (0.0%)	0/3 (0.0%)	NA	1/1 (100.0%)	0/1 (0.0%)	NA
CIPd	≥I +2	1/41 (2.4%)	NA	0/41 (0.0%)	7/48 (14.6%)	NA	0/48 (0.0%)
	I + 1 TO I-1	5/9 (55.6%)	0/9 (0.0%)	2/9 (22.2%)	0/2 (0.0%)	0/2 (0.0%)	0/2 (0.0%)
	≤I-2	0/0 (0.0%)	0/0 (0.0%)	NA	0/0 (0.0%)	0/0 (0.0%)	NA
PIP-TAZd	≥IHi +2	5/24 (20.8%)	NA	3/24 (12.5%)	2/25 (8.0%)	NA	2/25 (8.0%)
	IHi+1 TO ILo-1	16/19 (84.2%)	0/19 (0.0%)	1/19 (5.3%)	7/13 (53.8%)	0/13 (0.0%)	4/13 (30.8%)
	≤ILo-2	0/7 (0.0%)	0/7 (0.0%)	NA	0/12 (0.0%)	0/12 (0.0%)	NA
Acceptance criteria	≥I +2	<5%	NA	<2%	<5%	NA	<2%
	I + 1 TO I-1	<40%	<10%	<10%	<40%	<10%	<10%
	≤I-2	<5%	<2%	NA	<5%	<2%	NA

^aNA, not applicable.

^bBoldface indicates when acceptance criteria were not met.

^cDD zone diameters were interpreted using P. aeruginosa breakpoints.

^dBMD MICs and DD zone diameters were interpreted using P. aeruginosa breakpoints.

TABLE 7.

Comparison of MicroScan WalkAway to BMD using the error-rate bounded method by species

		B. cenocepacia			B. multivorans
Antibiotic	Type	mE	ME	VME	mE	ME	VME
CAZ	≥I +2	2/17 (11.8%) a	NAb	4/17 (23.5%)	0/20 (0.0%)	NA	4/20 (20.0%)
	I + 1 TO I-1	4/22 (18.2%)	0/22 (0.0%)	5/22 (22.7%)	2/16 (12.5%)	2/16 (12.5%)	3/16 (18.8%)
	≤I-2	0/11 (0.0%)	2/11 (18.2%)	NA	0/14 (0.0%)	4/14 (28.6%)	NA
MEM	≥I +2	3/11 (27.3%)	NA	8/11 (72.7%)	5/23 (21.7%)	NA	13/23 (56.5%)
	I + 1 TO I-1	13/35 (37.1%)	0/35 (0.0%)	16/35 (45.7%)	9/26 (34.6%)	0/26 (0.0%)	14/26 (53.8%)
	≤I-2	0/4 (0.0%)	0/4 (0.0%)	NA	0/1 (0.0%)	0/1 (0.0%)	NA
TMP-SMX	≥R +1	NA	NA	8/22 (36.4%)	NA	NA	5/22 (22.7%)
	R + S	NA	2/14 (14.3%)	8/14 (57.1%)	NA	1/12 (8.3%)	4/12 (33.3%)
	≤S -1	NA	2/14 (14.3%)	NA	NA	3/16 (18.8%)	NA
LVX	≥I +2	1/30 (3.3%)	NA	12/30 (40.0%)	5/27 (18.5%)	NA	5/27 (18.5%)
	I + 1 TO I-1	9/17 (52.9%)	0/17 (0.0%)	4/17 (23.5%)	13/22 (59.1%)	0/22 (0.0%)	3/22 (13.6%)
	≤I-2	0/3 (0.0%)	2/3 (66.7%)	NA	0/1 (0.0%)	0/1 (0.0%)	NA
PIP-TAZc	≥IHi +2	9/24 (37.5%)	NA	10/24 (41.7%)	6/25 (24.0%)	NA	8/25 (32.0%)
	IHi+1 TO ILo-1	15/19 (78.9%)	0/19 (0.0%)	3/19 (15.8%)	5/13 (38.5%)	0/13 (0.0%)	5/13 (38.5%)
	≤ILo-2	0/7 (0.0%)	1/7 (14.3%)	NA	0/12 (0.0%)	2/12 (16.7%)	NA
Acceptance criteria	≥I +2	<5%	NA	<2%	<5%	NA	<2%
	I + 1 TO I-1	<40%	<10%	<10%	<40%	<10%	<10%
	≤I-2	<5%	<2%	NA	<5%	<2%	NA

^aBoldface indicates when acceptance criteria were not met.

^bNA, not applicable.

^cBMD and MicroScan WalkAway MICs were interpreted using P. aeruginosa breakpoints.

Statistics.

To determine whether errors were enriched in either B. cenocepacia or B. multivorans, two-by-two contingency tables were created for each species with the numbers of agreeing results between methods or errors between methods. Fisher’s exact test was performed using Prism GraphPad software v9.1.2 using the VMEs, MEs, and mEs calculated from the analyses of DD and MicroScan WalkAway relative to BMD, using the mode values (or median values, when applicable) from each assay to determine agreement or errors, as described above.

RESULTS

Antimicrobial susceptibility.

BMD analyses were performed in triplicate to determine in vitro susceptibilities to the following drugs: CAM, CAZ, LVX, MEM, MIN, TMP-SMX, PIP-TAZ, and CIP. The percentages of isolates testing susceptible, intermediate, or resistant are shown in Fig. 1. Overall, the isolates showed low susceptibility rates to all drugs tested. CAZ and TMP-SMX had the highest in vitro activity, with 43% and 39% of isolates testing susceptible, respectively. CAM and CIP had the lowest in vitro activity, with only 1% of isolates testing susceptible. For PIP-TAZ, MIN, MEM, and LVX, 23%, 21%, 15%, and 9% of isolates tested susceptible, respectively.

FIG 1.

AST results for 100 CF BCC isolates. BMD was performed in triplicate on 50 B. multivorans and 50 B. cenocepacia isolates following CLSI guidelines. Data represent the mode for each isolate. If no mode was available, then the median value was used. MICs were interpreted as susceptible (S), intermediate (I), or resistant (R) using CLSI breakpoints for BCC, except for ciprofloxacin and piperacillin-tazobactam, where P. aeruginosa breakpoints were applied. Abbreviations: CAZ, ceftazidime; MEM, meropenem; MIN, minocycline; LVX, levofloxacin; CAM, chloramphenicol; CIP, ciprofloxacin; TMP-SMX, trimethoprim-sulfamethoxazole; PIP-TAZ, piperacillin-tazobactam.

Reproducibility.

BMD reproducibility results are shown in Table 2. CAZ, LVX, CAM, CIP, TMP-SMX, and PIP-TAZ did not meet acceptance criteria for reproducibility, with 82%, 93%, 94%, 94%, 88%, and 57% of MICs falling within ± 1 log₂ dilution between replicates. MEM (97%) and MIN (97%) showed acceptable reproducibility results. Similar to BMD, MicroScan WalkAway reproducibility was acceptable for MEM (100%) and unacceptable for PIP-TAZ (88%) (Table S1). However, CAZ (96%), LVX (100%), and CIP (100%) showed acceptable reproducibility. A technical limitation of the MicroScan WalkAway analysis was that more isolates were excluded than were from BMD because the MICs were not on scale. DD reproducibility is shown in Table 3. For replicates within 3 mm, DD showed low reproducibility for all drugs tested (64-86%). DD reproducibility improved for replicates within 6 mm and was > 90% for LVX (94%), TMP-SMX (95%), MIN (96%), and CIP (97%). However, reproducibility was still < 90% for PIP-TAZ (77%), CAZ (81%), and MEM (89%). At 3 mm and 6 mm, PIP-TAZ showed the lowest reproducibility (64% and 77%, respectively), while TMP-SMX and CIP had the highest reproducibility (86% and 95%; 83% and 97%, respectively).

TABLE 2.

Reproducibility of BMD for 100 BCC isolates tested on 3 consecutive days

Antibiotic	Same MIC and within ±1 log2	Same MIC	±1 log2 dilution	±2 log2 dilution	±3 log2 or more dilution
CAZ	82% (70/85)	36% (31/85)	46% (39/85)	11% (9/85)	7% (6/85)
MEM	97% (93/96)	59% (57/96)	38% (36/96)	3% (3/96)	0% (0/96)
MIN	97% (89/92)	52% (48/92)	45% (41/92)	2% (2/92)	1% (1/92)
LVX	93% (70/75) a	60% (45/75)	33% (25/75)	7% (5/75)	0% (0/75)
CAM	94% (59/63)	60% (38/63)	33% (21/63)	5% (3/63)	2% (1/63)
CIP	94% (62/66)	50% (33/66)	44% (29/66)	6% (4/66)	0% (0/66)
TMP-SMX	88% (73/83)	42% (35/83)	46% (38/83)	11% (9/83)	1% (1/83)
PIP-TAZ	57% (24/42)	33% (14/42)	24% (10/42)	24% (10/42)	19% (8/42)

^aBoldface indicates when acceptance criteria were not met.

TABLE 3.

Reproducibility of DD for 100 BCC isolates tested on 3 consecutive days

	DD range
Antibiotic	Within 3 mm	Within 6 mm
CAZ	67/100 (67%)	81/100 (81%)
MEM	71/100 (71%)	89/100 (89%)
MIN	73/100 (73%)	96/100 (96%)
LVX	83/100 (83%)	94/100 (94%)
CIP	83/100 (83%)	97/100 (97%)
TMP-SMX	86/100 (86%)	95/100 (95%)
PIP-TAZ	64/100 (64%)	77/100 (77%)

Comparison of DD and BMD.

Overall, DD performed sub-optimally for all drugs tested except MEM (Fig. 2A and Table 4). MEM was the only drug for which DD testing met acceptance criteria in all three categories of isolates. For TMP-SMX and LVX, mE rates were unacceptably high among susceptible isolates (TMP-SMX, n = 2/30, 6.7%; LVX n = 1/4, 25%). CAZ had unacceptable VME rates (n = 3/37, 8.1%). Finally, DD testing for MIN, PIP-TAZ, and CIP performed the most poorly, having unacceptable mE and VME rates. Only 1 ME was detected for TMP-SMX (n = 1/26, 3.8%), and this rate met the < 10% acceptance threshold for isolates in the intermediate susceptibility group. Generally, the errors observed tended to understate resistance.

FIG 2.

Scattergrams comparing DD and MicroScan WalkAway to BMD for 100 CF BCC isolates. Scattergrams were generated to compare mode or median BMD MIC data to mode or median DD zone diameter (A) and mode or median MicroScan WalkAway MIC (B). The number of isolates testing resistant (R), intermediate (I), or susceptible (S) by BMD is shown on each plot. BCC breakpoints were applied except for the following, where P. aeruginosa breakpoints were applied: LVX, CIP, and PIP-TAZ for DD; PIP-TAZ and CIP for BMD; PIP-TAZ for MicroScan WalkAway. For TMP-SMX and PIP-TAZ, only the TMP and PIP MICs are shown, respectively. Minor errors are indicated in yellow, major errors are indicated in orange, and very major errors are indicated in red. Solid lines within the scattergrams indicate breakpoints.

Comparison of MicroScan WalkAway and BMD.

Since DD accuracy was unacceptably low for all antibiotics tested except MEM, and the number of BCC isolates tested in previous performance studies for automated AST methods is low (23–25), we compared the accuracy of MicroScan WalkAway to BMD (Table 5). MicroScan WalkAway performed poorly for all antibiotics tested (Fig. 2B). VME rates were unacceptably high for all antibiotics; ME rates were unacceptably high for CAZ, TMP-SMX, LVX, and PIP-TAZ; and mE rates were unacceptable for CAZ, MEM, LVX, and PIP-TAZ. As observed for DD, the MicroScan WalkAway also tended to understate resistance.

Performance of DD and MicroScan WalkAway by BCC species.

To determine whether poor performance of DD and MicroScan WalkAway was species-specific, we compared DD and MicroScan WalkAway data to BMD for B. cenocepacia and B. multivorans (Tables 6 and 7, Fig. 3). When comparing DD to BMD, no significant differences were noted between B. cenocepacia and B. multivorans. When comparing MicroScan WalkAway to BMD, there were slightly more errors for PIP-TAZ for B. cenocepacia (Fig. 3B, P < 0.05, Fisher’s exact test). Overall, DD and MicroScan WalkAway performed comparably between B. cenocepacia and B. multivorans.

FIG 3.

Comparison of error rates for each antibiotic for DD and MicroScan WalkAway to BMD by species. (A) DD errors compared to BMD for each antibiotic. Table shows P values for Fisher’s exact test. (B) MicroScan WalkAway errors compared to BMD for each antibiotic. Table shows P values for Fisher’s exact test. Bcen: B. cenocepacia, Bmul: B. multivorans. Errors include VMEs, MEs, and mEs.

DISCUSSION

BCC organisms are notorious pathogens in individuals with CF because they can cause rapid decline in lung function and invasive disease. These infections are difficult to treat due to intrinsic and acquired resistance to antibiotics used in current practice (4, 7–14, 32–36). A final option for treatment is lung transplantation, which may prolong patient survival (3). To guide treatment and eligibility for lung transplant, AST results should be reproducible and accurate. In our study, BMD reproducibility was acceptable for MEM and MIN, but unacceptable for CAZ, LVX, CAM, CIP, TMP-SMX, and PIP-TAZ (Table 2). MicroScan WalkAway reproducibility was similar (Table S1). Wootton et al. found poor BMD reproducibility for CAM, CAZ, CIP, MEM, MIN, and TMP-SMX (20). In their study, MIN had the highest reproducibility at 84.5%, while the remaining drugs had reproducibilities between 70.3 and 79.3% (20). The reproducibility rates in this study were higher (57–97%), and one explanation may be differences in the brands of media that were used. However, similar to Wooton et al., most drugs did not meet the ≥ 95% cut-off that was used by both studies (20). DD reproducibility was poorer than BMD and MicroScan WalkAway, with all drugs tested < 90% reproducible when a 3 mm cut-off was applied. This finding is significant since 3 mm differences may impact interpretive category when reading DD zones of inhibition. Even when extending the cut-off to 6 mm, important drugs like MEM and CAZ still had low reproducibility (< 90%). Together, these data suggest that DD may not be reliably reproducible for AST of B. multivorans and B. cenocepacia, and BMD and MicroScan WalkAway are only reliably reproducible for some drugs. Notably, CAZ and TMP-SMX are two first-line drugs used for BCC treatment (28, 37), and both showed poor BMD reproducibility.

An informal survey of clinical microbiology laboratories found that 4/16 (25%) use DD testing for BCC isolates (unpublished data). Here we show that compared to BMD, DD performed sub-optimally for B. cenocepacia and B. multivorans. MEM was the only drug that passed all acceptance criteria. While TMP-SMX and LVX did not meet acceptance criteria due to high mE rates, these errors are not as concerning as VMEs and MEs. In contrast, CAZ had unacceptably high VMEs, while MIN, PIP-TAZ, and CIP had both unacceptable mE and VME rates, suggesting that DD results for these drugs are problematic because they under-call resistance. These findings align well with previous studies that compared performance of DD to BMD for BCC isolates. Acceptance criteria were not met for any drugs in the Wootton et al. study (20), while Fehlberg et al. found that all acceptance criteria were met for only MEM and MIN (21). One explanation for the poor performance of DD compared to BMD may be different growth rates of isolates. It was noted by Wootton et al. that particular isolates showed poor reproducibility and correlation for all methods, including growth at lower temperatures (20). In this study, it was noted that some isolates (n = 21) did not grow as robustly on standard growth media as other isolates, even though they met CLSI criteria for interpreting zones of inhibition and MIC. These isolates may have benefited from another 24 h of incubation or a different brand of media, which we are pursuing in future studies.

Some clinical microbiology laboratories may opt to perform automated AST methods for BCC isolates. Previous studies on the performance of automated AST methods for BCC AST tested only small numbers of isolates (23–25). In our study, MicroScan WalkAway performed more poorly than DD, with unacceptable results for all drugs. Testing produced unacceptably high VMEs for CAZ, MEM, TMP-SMX, LVX, and PIP-TAZ. Additionally, CAZ, LVX, and PIP-TAZ had unacceptably high MEs and mEs, while MEM had unacceptably high mEs. Similar to BMD and DD, isolate growth may have impacted performance. Additionally, MicroScan WalkAway is not FDA-cleared for reporting PIP-TAZ on BCC isolates, which may explain its poor performance. It remains that other automated AST platforms may differ in their performance, and future studies are required to fully evaluate their utility and accuracy.

We also explored the possibility that the poor correlation between AST methods could be caused by unique characteristics of each BCC species. We therefore compared AST results from DD and MicroScan WalkAway to BMD for B. cenocepacia (n = 50) and B. multivorans (n = 50) (Tables 6 and 7, Fig. 3). No significant differences were noted between species except for PIP-TAZ when tested by MicroScan WalkAway, although this difference was very slight. Similarly, Wootton et al. found that poor BMD reproducibility was not species-specific, although noted that there were low numbers for some species tested (20). Our findings, combined with results from the Wootton et al. and Fehlberg et al. studies, suggest that DD methods generally perform poorly for all BCC species (20, 21). While more isolates may need to be tested for all species within the complex, B. gladioli, and non-CF isolates to make these results more generalizable, the most common CF isolates have been tested in higher numbers among these studies (B. cepacia, n = 64; B. cenocepacia, n = 142; B. multivorans, n = 81).

There are some limitations to this study. Our isolates showed low susceptibility rates to all drugs tested (Fig. 1), and > 20% of the isolates had MIC values within ± 1 log₂ dilution of the CLSI breakpoints. Isolates at or near the breakpoints may account for some error between methods. However, Fehlberg et al. tested isolates that had higher susceptibility rates (except CAM) and found similar results to this study (21). Therefore, it seems likely that some AST methods may perform poorly for most BCC CF isolates, regardless of species and susceptibility. Technical limitations to the study include that P. aeruginosa breakpoints had to be applied for several drugs, MicroScan WalkAway was performed at a different site than BMD and DD, and we did not test media and reagents from multiple manufacturers. Finally, all the isolates from this study were collected from individuals with CF, where it is known that BCC bacteria can evolve and adapt to the CF lung environment (9, 10, 38, 39). These adaptations could affect the performance of AST reproducibility and correlation among methods. Slow growth is a common adaptation to the CF lung environment (38), and as mentioned previously, we noted throughout the study that some isolates grew more slowly than others. In addition to testing longer incubation times, we plan to evaluate different brands or types of growth media in future work as a way to improve AST for slow-growing CF BCC isolates. Finally, due to the unique characteristics of CF isolates, it will also be important to evaluate performance of AST methods for non-CF BCC isolates, particularly those from acute infections where adaptations are less likely to occur.

The results from this study generally support findings from EUCAST and previous studies showing that performance of AST methods for BCC is poor. In contrast to the Wootton et al. study (20), we found that BMD was reproducible for two antibiotics, suggesting that this method may still be reliable for some drugs; however, we observed low reproducibility for CAZ and TMP-SMX, two first-line drugs for treatment of BCC infections, and given the problems encountered with numerous AST methods, it is still not known whether BMD is a meaningful gold standard. The exception was MEM, which met acceptance criteria for BMD reproducibility and DD accuracy. Additionally, our data show that resistance is under-called by DD and MicroScan WalkAway, a potentially dangerous problem for patient treatment. In light of these findings, previous studies, EUCAST recommendations, and the analyses performed by the AMR in the CF International Working Group (19), routine AST for BCC CF isolates should be interpreted with caution. Long-term, establishing better ways to assess susceptibility in BCC isolates will be a critical step in improving CF care.