Evaluation of Ciprofloxacin and Levofloxacin Disk Diffusion and Etest Using the 2019 Enterobacteriaceae CLSI Breakpoints

In 2019, the Clinical and Laboratory Standards Institute (CLSI) published revisions to the Enterobacteriaceae ciprofloxacin and levofloxacin breakpoints. We evaluated the performance of disk diffusion and Etest compared to that of reference broth microdilution by use of the revised breakpoints.

ABSTRACT

In 2019, the Clinical and Laboratory Standards Institute (CLSI) published revisions to the Enterobacteriaceae ciprofloxacin and levofloxacin breakpoints. We evaluated the performance of disk diffusion and Etest compared to that of reference broth microdilution by use of the revised breakpoints. Fifty-eight Enterobacteriaceae isolates with ciprofloxacin MICs of 0.5 μg/ml or 1.0 μg/ml on initial testing were specifically selected for evaluation. These MICs are susceptible by the 2018 breakpoints and not susceptible by the 2019 breakpoints. For ciprofloxacin disk diffusion, the categorical agreement (CA) was 46.6%, with 0 very major errors (VME), 4 major errors (ME) (21.1%), and 27 minor errors (mE) (46.6%) using the 2019 CLSI disk breakpoints. For levofloxacin, the CA was 72.4%, with 0 VME, 0 ME, and 16 mE (27.6%) using the 2019 CLSI disk breakpoints. Using an error rate-bound evaluation method, levofloxacin but not ciprofloxacin disk diffusion yielded an acceptable minor error rate of <40% for isolates with an MIC plus or minus 1 doubling dilution of the intermediate breakpoint. For Etest compared to the reference broth microdilution, the essential agreement was 100% for both ciprofloxacin and levofloxacin and the CA was 81.0% and 65.5%, respectively. No VME or ME were observed by Etest, and 11 minor errors for ciprofloxacin (19.0%) and 20 (34.5%) for levofloxacin were observed. By the error rate-bound method, the minor error rate for ciprofloxacin was acceptable, but minor error rates for levofloxacin remained outside the acceptance range (i.e., 42.6% for isolates with an MIC within 1 dilution of the breakpoint). In general, the disk diffusion and Etest methods performed well with this challenging collection of isolates, although laboratories must be aware of minor errors, particularly for isolates with results near the breakpoint.

INTRODUCTION

In 2018, the Clinical and Laboratory Standards Institute (CLSI) approved a revision to the Enterobacteriaceae and Pseudomonas aeruginosa ciprofloxacin and levofloxacin MIC and disk diffusion (DD) breakpoints. The revised breakpoints were published in the 29th edition of the M100 Supplemental Performance Standards for Antimicrobial Susceptibility Testing in January 2019 (1) and are the same as those published by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (see Table 2). The U.S. FDA has not yet recognized the revised fluoroquinolone breakpoints, although CLSI has submitted to the FDA a rationale document that outlines the data supporting the change for its review for potential recognition. Commercial antimicrobial susceptibility test (cAST) manufacturers may obtain FDA clearance for their systems only using FDA-recognized breakpoints. As such, they can revise the breakpoints applied to their cAST systems only after (or if) FDA recognizes the revised ciprofloxacin and levofloxacin CLSI breakpoints. Furthermore, manufacturers may need to reformulate the ciprofloxacin and/or levofloxacin on their cAST systems to accommodate the revised breakpoints, either because the dilution range for these drugs on their present systems is not low enough to accommodate the revised breakpoints or because the system performance is found to be less accurate with the revised breakpoints. In these cases, the cAST manufacturers will need to perform a new clinical trial to assess the performance of their revised cAST systems with the revised breakpoints.

TABLE 2.

Ciprofloxacin and levofloxacin Enterobacteriaceae breakpoint summary^a

Susceptibility	Breakpoint for:
	Ciprofloxacin					Levofloxacin
	MIC (μg/ml)			Disk diam (mm)		MIC (μg/ml)			Disk diam (mm)
	CLSI 2018	EUCAST 2019	CLSI 2019	CLSI 2019	EUCAST 2019	CLSI 2018	EUCAST 2019	CLSI 2019	CLSI 2019	EUCAST 2019
S	≤1	≤0.25	≤0.25	≥26	≥26	≤2	≤0.5	≤0.5	≥21	≥23
I	2	0.5	0.5	22–25	24–25	4	1	1	17–20	19–22
R	≥4	>0.5	≥1	≤ 21	<24	≥ 8	>1.0	≥2	≤16	<19

^aCLSI, Clinical and Laboratory Standards Institute; EUCAST, European Committee on Antimicrobial Susceptibility Testing; S, susceptible; I, intermediate; R, resistant.

In any scenario, it is certain that laboratories will face a delay between the CLSI publication of the revised fluoroquinolone breakpoints and the availability of these on FDA-cleared cAST systems. Prior experience with the monobactam cephalosporin and carbapenem breakpoint revisions suggests that this delay may be several years, even after FDA’s recognition of the revised breakpoints. As such, laboratories will have to develop alternative susceptibility testing methods if they desire to adopt the revised ciprofloxacin and levofloxacin breakpoints. The aims of this study were to (i) assess the reproducibility of reference broth microdilution (BMD) MICs for isolates with MICs in the range of 0.5 to 1.0 μg/ml and (ii) evaluate the agreement of the disk diffusion and Etest methods with BMD categorical interpretations and essential agreement (EA) (Etest only) using the 2019 CLSI breakpoints.

MATERIALS AND METHODS

Isolates.

A collection of 58 Enterobacteriaceae isolates, including 20 Escherichia coli, 20 Klebsiella pneumoniae, 7 Proteus mirabilis, 4 Serratia marcescens, 2 Klebsiella oxytoca, and 2 Enterobacter aerogenes isolates, 1 Raoultella ornithinolytica isolate, 1 Kluyvera ascorbata isolate, and 1 Enterobacter cloacae isolate, were included in this study (Table 1). All had ciprofloxacin MICs of 0.5 to 1.0 μg/ml (i.e., the isolates were susceptible by the 2018 breakpoints and intermediate or resistant by the 2019 breakpoints) upon initial testing by the clinical laboratory, which was done by the reference broth microdilution (BMD) method. All isolates were recovered from the blood of unique patients at UCLA between 2015 and 2016. Isolates were stored at room temperature on tryptic soy agar slants (Hardy Diagnostics, Santa Ana, CA) overlaid with sterile mineral oil and subcultured twice on 5% sheep’s blood agar plates (BAP; BBL, BD, Sparks, MD) prior to testing.

TABLE 1.

Enterobacteriaceae isolates included in this study

Antimicrobial and organism	No. of isolates with reference MIC (μg/ml) of:
	0.03	0.06	0.12	0.25	0.5	1	2	4	Total
Ciprofloxacin
E. aerogenes				1		1			2
E. cloacae					1				1
E. coli				7	3	10			20
K. ascorbata							1		1
K. oxytoca				1		1			2
K. pneumoniae	2		3	2	10	3			20
P. mirabilis		1			2	4			7
R. ornithinolytica						1			1
S. marcescens		2			2				4
Levofloxacin
E. aerogenes						2			2
E. cloacae					1				1
E. coli					1	7	12		20
K. ascorbata							1		1
K. oxytoca				1	1				2
K. pneumoniae		4		1	10	5			20
P. mirabilis			1	1		5			7
R. ornithinolytica					1				1
S. marcescens			2		2				4

Susceptibility testing.

Disk diffusion (DD) was performed according to CLSI M02 standards (2), and Etest (bioMérieux, Durham, NC) was performed according to the manufacturer’s instructions. Briefly, a suspension of the test organism equivalent to a 0.5 McFarland standard was prepared in 0.85% saline from three to five well-isolated colonies from an overnight culture on BAP. Using a swab, the organism was inoculated onto a Mueller-Hinton agar (MHA; BBL, BD) plate. For each MHA plate, a 5-μg ciprofloxacin disk, a 5-μg levofloxacin disk, a ciprofloxacin Etest, and a levofloxacin Etest were applied. The disks were purchased from BD. The plates were incubated for 16 to 18 h at 35°C, and zones were read visually, using reflected light. If discrete colonies were present within the zone of growth inhibition or abnormal growth was present, the colony was subcultured and checked for purity on a BAP.

Isolates were tested in parallel, using the same inoculum suspension, by the CLSI reference BMD method for ciprofloxacin and levofloxacin MICs on panels prepared in-house according to CLSI M07 standards (3) using 2 lots of stock solution for ciprofloxacin and levofloxacin and 2 manufacturers’ broths (Difco, BD, and BBL, BD). In this manner, 4 replicate ciprofloxacin and levofloxacin BMD MICs were obtained for each isolate. Ciprofloxacin and levofloxacin powders were obtained from Sigma, and the ciprofloxacin and levofloxacin concentrations in the BMD panels ranged from 0.03 μg/ml to 8 μg/ml. The panels were incubated for 16 to 20 h at 35°C, and any abnormal growth pattern observed in the BMD panels was subcultured to confirm the purity, in addition to performance of a routine purity plate. Quality control (QC) was assessed using Escherichia coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 for DD, Etest, and BMD (4).

Data analysis.

The BMD MIC mode, calculated from the 4 replicate BMD results, was used as the reference MIC. For Etest comparison, MICs were rounded up to the nearest 2-fold dilution for comparison to the BMD reference value. The BMD reference MIC for each organism was compared against the DD zone of inhibition diameter and Etest MIC. Disk and MIC results were evaluated by the use of 2018 (3) and 2019 (1) CLSI breakpoints and the EUCAST disk breakpoints (Table 2). Very major errors (VME), major errors (ME), and minor errors (mE) were calculated as described elsewhere (5) and also by the error rate-bound method (5). Overall categorical agreement (CA) with the BMD mode was evaluated for the disk and Etest methods, and essential agreement (EA) with the BMD mode was calculated for Etest.

RESULTS

Precision of fluoroquinolone MICs for isolates with ciprofloxacin MICs in the 0.5- to 1.0-μg/ml range.

The majority of isolates (50/58) yielded an MIC that was within 1 2-fold dilution of the initial results obtained by the clinical laboratory during standard-of-care testing. For eight isolates, the BMD MICs were on the low end of the susceptible range (i.e., 0.06 to 0.12 μg/ml), which reproduced upon repeat BMD testing. The majority of isolates yielded one (51.7%) or two (46.6%) MIC values across the four replicate ciprofloxacin MIC tests per reference BMD panel. Similarly, 43.1% of isolates yielded one MIC result and 53.5% yielded two MIC results for levofloxacin, indicating good reproducibility across the replicate tests. No trend was found between MICs obtained by the Difco versus BD brand of cation-adjusted Mueller-Hinton broth (not shown). All QC results were within CLSI published ranges (not shown).

Performance of disk diffusion.

The performance of the ciprofloxacin and levofloxacin disk diffusion tests compared to the modal BMD MICs is shown in Fig. 1. For ciprofloxacin, CA was 46.6%, with 0 VME, 4 ME (21.1%), and 27 mE (46.6%), using the 2019 CLSI disk breakpoints. For levofloxacin, CA was 72.4%, with 0 VME, 0 ME, and 16 mE (27.6%), using the 2019 CLSI disk breakpoints. Among the minor errors observed for the ciprofloxacin disk, 23 (85.2%) were due to the disk result being interpreted in a more resistant category (i.e., 9 were susceptible by MIC and intermediate by the disk and 14 were intermediate by MIC and resistant by the disk). For levofloxacin, 9 minor errors were due to a susceptible MIC and an intermediate disk result, and 7 were due to an intermediate MIC and a susceptible disk result (Fig. 1). When the data were evaluated by using EUCAST disk breakpoints, CA was 46.6% for ciprofloxacin, with 0 VME, 7 (36.8%) ME, and 24 (41.4%) mE. Most minor errors by the EUCAST interpretations were due to disk results being interpreted in a more resistant category (22 one category more susceptible by the MIC than by the disk versus 2 more resistant by the MIC). CA was 56.9% for levofloxacin, with 0, 1 (3.0%), and 24 (41.4%) VME, ME, and mE, respectively. Again, most minor errors by the EUCAST interpretations were due to disk results being interpreted in a more resistant category than the MIC (22 more resistant by the disk and 2 more resistant by the MIC).

FIG 1.

Disk diffusion diameter-to-MIC scattergrams for 58 Enterobacteriaceae isolates evaluated by use of the 2019 CLSI breakpoints (black lines). EUCAST disk breakpoints are depicted by the red lines.

Because the isolates evaluated were enriched for those with MICs outside the wild-type MIC, the data were evaluated using the error rate-bound method recommended by CLSI. By this method, the inherent error of the MIC method (i.e., plus or minus a single doubling dilution) is taken into account, with a higher minor error rate being allowable for isolates with MICs a single doubling dilution from the intermediate breakpoint (Table 3). The results of this analysis are shown in Table 4 for CLSI breakpoints. The disk diffusion test met the acceptance criteria for levofloxacin but not for ciprofloxacin, where the minor error rate was 49.0%, higher than the 40% acceptance limit published in the M23 guideline (5).

TABLE 3.

Error rate-bound method acceptance limits according to CLSI M23 (5) guidance^a

MIC range	% of the following:
	VME	ME	mE
I + ≥2	<2	NA	<5
I ± 1	<10	<10	<40
I − ≤2	NA	<2	<5

^aCLSI, Clinical and Laboratory Standards Institute; I + ≥2, MIC greater than or equal to 2 doubling dilutions of the intermediate breakpoint; I ± 1, MIC plus or minus 1 doubling dilution of the intermediate breakpoint intermediate category; I − ≤2, MIC less than or equal to 2 doubling dilutions of the intermediate breakpoint; VME, very major error; ME, major error; mE, minor error; NA, not applicable.

TABLE 4.

Error rate-bound evaluation of ciprofloxacin and levofloxacin disk diffusion and Etest for a collection of Enterobacteriaceae^a

MIC range and antimicrobial	No. of isolates	No. (%) of isolates with the indicated error by the following method:
		CLSI disk			EUCAST disk			Etest
		VME	ME	mE	VME	ME	mE	VME	ME	mE
Ciprofloxacin
I + ≥2 μg/ml	1	0 (0)		0 (0)	0 (0)		0 (0)	0 (0)		0 (0)
I ± 1 μg/ml	49	0 (0)	4 (8.16)	24 (49.0)		7 (14.3)	21 (42.8)	0 (0)	0 (0)	11 (22.4)
I − ≤2 μg/ml	8		0 (0)	3 (37.5)		0 (0)	3 (37.5)		0 (0)	0 (0)
Levofloxacin
I + ≥2 μg/ml	0
I ± 1 μg/ml	47	0 (0)	0 (0)	16 (34.0)	0 (0)	1 (2.1)	24 (51.1)	0 (0)	0 (0)	20 (42.6)
I − ≤2 μg/ml	11		0 (0)	0 (0)		0 (0)	0 (0)		0 (0)	0 (0)

^aData are for 58 isolates. CLSI, Clinical and Laboratory Standards Institute; EUCAST, European Committee on Antimicrobial Susceptibility Testing; I + ≥2, MIC greater than or equal to 2 doubling dilutions of the intermediate breakpoint; I ± 1, MIC plus or minus 1 doubling dilution of the intermediate breakpoint intermediate category; I − ≤2, MIC less than or equal to 2 doubling dilutions of the intermediate breakpoint; VME, very major error; ME, major error; mE, minor error. Results in bold exceed acceptance standards according to the CLSI M23 method (5).

Performance of Etest with revised breakpoint.

The performance of Etest compared to the modal MIC from BMD is shown in Fig. 2. EA was 100% for both ciprofloxacin and levofloxacin, but CA was 81.0% and 65.5%, respectively. No VME or ME were observed by either the ciprofloxacin or the levofloxacin Etest. Minor errors were high, with 11 (19.0%) being observed for ciprofloxacin and 20 (34.5%) being observed for levofloxacin. The majority of minor errors were due to Etest calling a single MIC above the modal MIC from BMD (10/11 for ciprofloxacin and 18/20 for levofloxacin). An evaluation of the data using the error rate-bound method is shown in Table 4. By this analysis, the minor error rate for isolates with Etest MICs within a single dilution of the breakpoint was only 22.4%, acceptable by CLSI standards, but levofloxacin minor error rates remained outside the acceptance range (i.e., 42.6%).

FIG 2.

Etest MIC-to-BMD modal MIC scattergrams for 58 Enterobacteriaceae isolates evaluated by use of the 2019 CLSI breakpoints (black lines).

DISCUSSION

Since the ciprofloxacin and levofloxacin breakpoints were established several decades ago (6, 7), new tools have become available for examining pharmacokinetic and pharmacodynamic data and significant numbers of resistant strains have been encountered. Consequently, a need to reevaluate the breakpoints for these agents became apparent to CLSI. When doing so, pharmacokinetic and pharmacodynamic data demonstrated poor target attainment for the fluoroquinolones when the 2018 breakpoints for the Enterobacteriaceae were applied, and limited clinical outcome data demonstrated that these original breakpoints may not accurately predict the probability of treatment success (8). The CLSI Antimicrobial Susceptibility Testing Subcommittee thus approved a revision to these breakpoints for publication in 2019 (1). The updated ciprofloxacin and levofloxacin breakpoints bring the CLSI MIC standards into concordance with those of EUCAST (Table 1). However, the disk breakpoints between these two organizations differ, even though both recommend use of a 5-μg ciprofloxacin disk and a 5-μg levofloxacin disk (Table 1). CLSI disk diffusion breakpoints were established using data evaluated by EUCAST for establishment of its breakpoint, supplemented with the data presented in this work to yield a sufficient number of isolates required to set breakpoints according to the CLSI M23 standard (5). While no disk breakpoint yielded excellent performance, even with the expanded group of isolates, the breakpoints chosen by CLSI minimized VME and ME. While mE rates were high, they achieved the acceptance limit of the error rate-bound method, albeit through the use of a broader levofloxacin-intermediate zone. CLSI uses data from quality control studies to determine the relative variability of disk diffusion results as an aid for determining a reasonable intermediate range. A rule of thumb is to consider an intermediate category that spans no less than one-half the QC range and no more than the actual QC range. For E. coli ATCC 25922, the QC range for ciprofloxacin and levofloxacin disks is 29 to 37 mm (an 8-mm range). As such, an intermediate range of 4 to 8 mm is considered reasonable and in accordance with the 2019 CLSI intermediate breakpoints, which are 4 mm. When evaluating the data set presented herein, laboratories must be aware that the set of isolates tested was a challenge set for which MICs were very near the new breakpoints. As such, for the general population of Enterobacteriaceae tested, the performance of disk diffusion and Etest would be much better as a whole across all isolates tested by a laboratory.

The FDA has not yet approved the revised ciprofloxacin and levofloxacin breakpoints, meaning that antimicrobial susceptibility test (AST) manufacturers cannot yet obtain FDA clearance for their systems with the revised breakpoints. Laboratories that desire to adopt the revisions will need to perform a validation study of their AST systems to confirm that their performance is adequate, presuming that the laboratory uses an AST device that can accommodate the revised breakpoints. Many systems in the United States test dilutions no lower than 1 μg/ml for ciprofloxacin and 2 μg/ml for levofloxacin (S. Butler-Wu, submitted for publication). As a result, laboratories may need to adopt manual methods to determine the MIC for those isolates with ciprofloxacin MICs of ≤1 μg/ml or levofloxacin MICs of ≤2 μg/ml (which could fall into the susceptible, intermediate, or resistant interpretive categories by use of the revised breakpoints). Data from the SENTRY collection demonstrated that for isolates collected from 2011 to 2013, <4% of Enterobacteriaceae had MICs of 0.5 to 1 μg/ml for ciprofloxacin or 1 to 2 μg/ml for levofloxacin (susceptible by the 2018 breakpoint and not susceptible by the 2019 breakpoint). However, in the same collection, 81% of Enterobacteriaceae had a ciprofloxacin MIC of ≤1 μg/ml and 82.3% of isolates had a levofloxacin MIC of ≤2 μg/ml, which could be susceptible, intermediate, or resistant by use of the revised breakpoints. As a result, laboratories would need to test nearly all isolates by manual methods (the gradient diffusion or disk method) to accommodate the 2019 breakpoints. An alternative is validation of commercial AST systems with the 2019 breakpoints, if these systems have dilutions of antimicrobials that are low enough to accommodate the breakpoint. Unfortunately, fluoroquinolones are workhorse antimicrobials in hospitals, making the lack of AST systems FDA cleared with the revised breakpoints challenging. Furthermore, as is demonstrated by this study, the frequency of minor errors (i.e., overcalling resistance) is high by both the disk and Etest methods. While it is to be expected that AST performance will be poor when testing a challenging, non-wild-type organism collection, such as the one evaluated in this study, laboratories need to be aware that a trend toward more resistant results may be obtained by these methods.

Laboratories should discuss the revised breakpoints and the testing limitations with the infectious diseases physicians, pharmacists, antimicrobial stewardship program members, and any other invested parties at the facilities that they serve. Alternative strategies, such as selecting certain patient populations/clinical services or isolates recovered from select anatomical sites for testing by manual methods, may be a stopgap until a cleared AST device is available or the laboratory can validate its test system for use with the 2019 breakpoints. As an example, isolates from patients with uncomplicated urinary tract infections may not need testing by use of the 2019 breakpoints.

A limitation of this study is that only one brand each of MHA (BD), disks (BD), and gradient diffusion strips (Etest) was used. Results from this study may not apply to other brands of media, disks, or strips, and laboratories should carefully evaluate these if used. Similarly, we evaluated only 58 isolates because only ∼4% of Enterobacteriaceae isolates have MICs in the range that we desired to evaluate. It should be noted that the historical fluoroquinolone disk diffusion breakpoints were established using a set of isolates with a very limited number of resistant strains, 9 for ciprofloxacin (7) and 31 total (Gram-positive and Gram-negative strains) for levofloxacin (6). In summary, we demonstrated that the ciprofloxacin and levofloxacin disk and Etest methods perform reasonably well with the 2019 CLSI breakpoints, but laboratories should be cognizant of minor errors, particularly for isolates with disk or Etest results near the breakpoint. In such cases, discussion with the pharmacist or treating physician may be warranted.