The Resistant-Population Cutoff (RCOFF): a New Concept for Improved Characterization of Antimicrobial Susceptibility Patterns of Non-Wild-Type Bacterial Populations

Abstract

This study aimed to determine resistant-population cutoffs (RCOFFs) to allow for improved characterization of antimicrobial susceptibility patterns in bacterial populations. RCOFFs can complement epidemiological cutoff (ECOFF)-based settings of clinical breakpoints (CBPs) by systematically describing the correlation between non-wild-type and wild-type populations. We illustrate this concept by describing three paradigmatic examples of wild-type and non-wild-type Escherichia coli populations from our clinical strain database of disk diffusion diameters. The statistical determination of RCOFFs and ECOFFs and their standardized applications in antimicrobial susceptibility testing (AST) facilitates the assignment of isolates to wild-type or non-wild-type populations. This should improve the correlation of in vitro AST data and distinct antibiotic resistance mechanisms with clinical outcome facilitating the setting and validation of CBPs.

INTRODUCTION

In antimicrobial susceptibility testing (AST), clinical breakpoints (CBPs) are set with the intent of predicting the clinical outcome (1–3). Epidemiological cutoffs (ECOFFs), as defined by the European Committee for Antimicrobial Susceptibility Testing (EUCAST), separate wild-type from non-wild-type populations, reviving the concept of microbiological breakpoints (4–9). EUCAST uses the ECOFF as one of several tools in the process of CBP setting. Different methods have been suggested to determine ECOFFs (5, 10).

For a clinical isolate, prediction of treatment success is usually based on a single laboratory AST result, which is subsequently classified according to the corresponding CBPs. However, both MICs and disk diffusion diameters show significant variations when tested repeatedly (1, 6). For wild-type populations, these variations have been shown to result from measurement imprecision and methodological and biological factors (5, 11, 12). As it is based on a single measurement, the true position of an isolate within a disk diameter or MIC distribution can only be approximated with a certain probability. In addition, it is unclear to what extent the in vitro susceptibility of a clinical isolate reflects the in vivo situation, as in vivo variations within a population are likely to be more pronounced than those observed in vitro in the standardized setting of AST.

CBPs should preferably avoid splitting the wild-type population into different interpretative categories as the wild-type is regarded to be a genotypic entity despite population-intrinsic (micro-)variations in drug susceptibility (6). This principle of not splitting wild-type populations by CBPs is widely acknowledged by EUCAST guidelines (13). By logical consequence, these considerations should similarly apply to non-wild-type populations, which represent distinct genotypic/phenotypic entities (“resistotypes”). Thus, CBP setting should not only avoid splitting the wild-type into different interpretative categories but also CBPs should avoid splitting a resistotype into different clinical categories as far as possible in order to minimize categorization errors and to improve reproducibility of AST reports (14). If this principle is acknowledged, cutoff values are needed to define the borders of non-wild-type populations representing a particular resistotype. Here, we propose the term “resistant-population cutoff” (RCOFF), i.e., the largest inhibition zone diameter (or the lowest MIC) delineating a non-wild-type population (Fig. 1). The aim of this study was to evaluate the RCOFF concept in various settings regarding the relative position of wild-type and non-wild-type diameter distributions and the implications for the definition of CBPs.

FIG 1.

Schematic representation of wild-type and non-wild-type populations and ECOFF and RCOFF values. The diagram illustrates an ideal inhibition zone diameter distribution. The wild-type population (black bars), i.e., isolates devoid of acquired resistance mechanisms to the drug studied, is bordered by the epidemiological cutoff (ECOFF, green line) at the lowest inhibition zone diameter. Two non-wild-type populations (“resistotypes”) are distinguishable (gray bars). A resistant-population cutoff (RCOFF, red line) located at the largest inhibition zone diameter can, in principle, be defined for each non-wild-type subpopulation (resistotype).

MATERIALS AND METHODS

Clinical isolates.

Clinical strains of Escherichia coli included in this study were isolated from 2010 until 2014 in the clinical laboratory of the Institute for Medical Microbiology at the University of Zurich. Only nonduplicate isolates considered clinically relevant were included, and the numbers for specific species-drug combinations are indicated in Table 1. Different wild-type isolate numbers reflect missing values for certain antibiotic-isolate combinations.

TABLE 1.

Cutoff values determined for E. coli populations analyzed in this study and three clinically relevant beta-lactam antibiotics, compared to the EUCAST ECOFF values and clinical breakpoints^a

Antibiotic	Resistotype/phenotype	No. of isolates	ECOFF (mm)	RCOFF (mm)	EUCAST ECOFF (mm)	EUCAST S CBPb (≥, mm)	EUCAST R CBP (<, mm)
Ceftriaxone	ESBL	901		23		23	20
	Wild type	4,781	25		25
Imipenem	ESBL	900		38		22	16
	Wild type	4,783	23		24
Amoxicillin-clavulanic acid	ESBL	901		26		19	19
	Wild type	4,782	18		19

^aEpidemiological cutoff (ECOFF) and resistant-population cutoff (RCOFF) values were determined on the basis of the ROC curves for E. coli ESBL and wild-type populations and three clinically relevant beta-lactam antibiotics. The EUCAST ECOFF values are reported as described elsewhere (3, 13).

^bCBP, clinical breakpoint.

Antibiotic susceptibility testing.

The AST of clinical isolates was performed by the disk diffusion method following standard procedures (15, 16). Müller-Hinton agar and antibiotic discs were obtained from Becton-Dickinson (Franklin Lakes, NJ) and i2a (Montpellier, France). The recording of the resulting inhibition zone diameters was performed with the Sirweb/Sirscan system (i2a). Phenotypic screening for extended-spectrum beta-lactamases (ESBLs) was performed with cefpodoxime (10 μg/disk) as the marker, as suggested by the EUCAST guidelines (17). Phenotypic and molecular ESBL confirmation was carried out as previously described (18). E. coli wild-type beta-lactam susceptibility was defined as isolates being wild type to ampicillin and wild type to cefoxitin (according to the EUCAST ECOFFs) and wild type to cephalothin (in-house eyeball ECOFF of ≥10 mm) to exclude isolates with broad-spectrum beta-lactamases, cephalosporinases, and porin alterations (14).

Statistical analysis and determination of cutoffs.

The cutoffs separating the wild-type and non-wild-type (ESBL) E. coli populations were determined by using a receiver operating characteristic (ROC) curve methodology, as described in an accompanying paper by Valsesia et al. (19). ROC curves were generated for the ESBL/wild-type E. coli populations with respect to the inhibition zone diameters for ceftriaxone, imipenem, and amoxicillin-clavulanic acid separately. The ECOFFs and RCOFFs were derived from the coordinates of the ROC curves, as the cutoffs corresponding to a specificity or sensitivity level of ≥99%, respectively.

Software.

Data were coded in Microsoft Excel 2010 software (Microsoft Corporation, Redmond, VA) and analyzed using SPSS Statistics software version 22 (IBM Corp., Armonk, NY).

RESULTS

Inhibition zone diameter distributions of bacterial populations are characterized by the presence of a wild-type population and several non-wild-type populations (resistotypes), which carry different resistance mechanisms. As illustrated in Fig. 1, the lowest inhibition zone diameter of the wild-type population corresponds to the ECOFF. Several RCOFFs can be defined, each of which corresponds to a specific resistotype and is located at the largest inhibition zone diameter. While this illustrative example shows a situation of clear separation between the wild-type population and the resistotypes, different distribution patterns are frequently observed.

Three distinct exemplary settings with respect to the relative position of wild-type and non-wild-type populations were selected for this study, on the basis of inhibition zone diameter distributions of clinical E. coli isolates and three clinically relevant beta-lactam antibiotics (ceftriaxone, imipenem, and amoxicillin-clavulanic acid). The settings chosen illustrate different implications for CBP definition and the correct categorization of a clinical isolate.

Setting 1: Separated wild-type and non-wild type populations.

The diameter distribution of the E. coli ESBL and E. coli wild-type populations for the third-generation cephalosporin ceftriaxone illustrates a setting with two clearly separated populations (Fig. 2A). In this specific setting, the resistance mechanism is highly effective against the antibiotic, and the statistically derived RCOFF (23 mm) is lower than the ECOFF (25 mm). A susceptible CBP of ≥23 mm is recommended by the current EUCAST guidelines, exactly matching the statistically derived RCOFF, while the EUCAST ECOFF of 25 mm is identical to our ROC curve-based value (Table 1). In this case, microbiological information inherent in the ECOFF and RCOFF values can be integrated with pharmacokinetic and pharmacodynamic (PK/PD) data and the clinical data for the determination of the corresponding resistant/susceptible CBP.

FIG 2.

(A) Setting 1: separated wild-type and non-wild-type populations. (B) Setting 2: coinciding wild-type and non-wild-type populations. Wild-type and non-wild-type (ESBL-positive) populations of clinical E. coli isolates: ceftriaxone (A); imipenem (B). Statistically derived ECOFFs (green lines) and RCOFFs (red lines), the EUCAST clinical breakpoints (CBPs) (susceptible CBP [orange lines] and resistant CBP [blue lines]), and the EUCAST ECOFFs (black lines) are shown. Panel A (ceftriaxone) illustrates well-separated wild-type and ESBL E. coli populations, i.e., a situation in which the acquired resistance mechanism confers antibiotic resistance. Panel B (imipenem) illustrates coinciding wild-type and ESBL E. coli populations, i.e., a situation in which the acquired resistance mechanism does not affect antibiotic susceptibility.

Setting 2: Coinciding wild-type and non-wild-type populations.

The diameter distribution of the E. coli ESBL and E. coli wild-type populations for imipenem illustrates a setting of coinciding populations (Fig. 2B). The smallest diameter found among the ESBL isolates corresponds to the smallest zone size of the wild-type population, and the RCOFF (38 mm) is considerably higher than the ECOFF (23 mm). Note that the ROC curve-based ECOFF is slightly different from the EUCAST ECOFF (Table 1). This is an example of a drug resistotype where the underlying resistance mechanism does not affect the specific antibiotic. In agreement with this observation, clinical guidelines recommend the use of carbapenems for treatment of serious infections due to ESBL-positive E. coli isolates (20, 21). However, the emergence of strains resistant to carbapenems has triggered the search for alternative therapeutic options, such as beta-lactam-inhibitor combinations (22, 23).

Setting 3: Overlapping wild-type and non-wild type populations.

The zone diameter distributions of the E. coli ESBL and E. coli wild-type populations for amoxicillin-clavulanic acid depict a situation where wild-type and non-wild-type populations are partially overlapping (Fig. 3). The therapeutic value of beta-lactam-inhibitor combinations for treating infections caused by ESBLs is still controversial, and different studies have led to contradictory data (14, 22, 24–28). Consequently, in such a setting, the efficacy of an antibiotic is questionable, and the risk of in vivo resistance development under therapy may not be neglected, regardless of the clinical classification of a strain (20, 28, 29).

FIG 3.

Setting 3: overlapping wild-type and non-wild-type populations. Overlapping wild-type and non-wild-type (ESBL-positive) populations of clinical E. coli isolates: amoxicillin-clavulanic acid. Statistically derived ECOFFs (green lines) and RCOFFs (red lines), the EUCAST clinical breakpoints (CBPs) (susceptible CBP [orange lines] and resistant CBP [blue lines] and the EUCAST ECOFFs are shown.

In this particular setting, the RCOFF (26 mm) is notably higher than the ECOFF (18 mm), and the zone of overlap between the ESBL and wild-type populations is substantial (9 mm) (Fig. 3). A large number of non-wild-type isolates (n = 395, 43.8% of all ESBLs) with a high risk of clinical misclassification are situated in the zone of overlap. The EUCAST guidelines recommend a susceptible/resistant CBP for amoxicillin-clavulanic acid of ≥/<19 mm, which is identical to the EUCAST ECOFF value (3, 13). By applying the EUCAST breakpoints to our zone diameter distribution for the E. coli ESBL/wild-type population, a large number of ESBL isolates (n = 302, 33.5%) would be classified as susceptible despite the presence of a resistance mechanism, while some wild-type isolates would be classified as resistant (n = 61, 1.3%). As recently published by EUCAST, additional outcome data supporting the breakpoints for this species-drug combination are required (29). Thus, until conclusive evidence is provided, the probability of the occurrence of very major errors and major errors as a consequence of misclassification should not be neglected (14).

In an overlapping setting, three options exist for the process of CBP definition to minimize clinical misclassification of wild-type and non-wild-type isolates: (i) implementation of a large intermediate zone corresponding to the zone of overlap, which would contain a high number of clinical isolates (in this paradigmatic case, 4,671 isolates [82% of all isolates]) and which would have no practical value; (ii) inference of the genotype of isolates in the overlapping zone by interpretative reading; and (iii) determination of the genotype by means of phenotypic and/or molecular methods.

DISCUSSION

Currently, the ECOFF is ambiguously defined by EUCAST both as the cutoff “separating the wild type from any population with at least one acquired resistance mechanism” or as “the upper MIC value of the wild-type distribution” (10, 13). However, for those drug-pathogen combinations in which the non-wild-type and the wild-type populations overlap, determination of the ECOFF is not feasible if the first definition is applied. By applying the second definition, an ECOFF which corresponds to the value of the smallest diameter of the wild-type population can be determined. To do so, however, prior classification of isolates as wild-type or non-wild-type is required, either by interpretative reading or phenotypic and/or molecular confirmation (30). Similar considerations apply to the RCOFF, which can only be determined if isolates have been previously assigned to the wild-type or non-wild-type population, depending on the presence or absence of a known resistance mechanism.

Analysis of antimicrobial susceptibility patterns based both on the ECOFFs and RCOFFs allows for a better description of resistotype populations and wild-type populations and may help to standardize the process of CBP definition. If populations are clearly separated, and the RCOFF is lower than the ECOFF (Fig. 2A), or if the wild-type and non-wild-type populations are coinciding (Fig. 2B), a clear-cut susceptible/resistant CPB, which is based on the ECOFF as well as on PK/PD data and on clinical outcome data, can be set. If non-wild-type and wild-type populations overlap (Fig. 3), determination of ECOFFs according to the current EUCAST definition is challenging (13). Statistical analysis of populations carrying a resistance mechanism (“positive” populations) together with wild-type populations (“negative” populations) allows determination of the lower end of the wild-type diameter distribution (i.e., the ECOFF) as well as the upper end of the non-wild-type population (i.e., the RCOFF), as described in an accompanying paper by Valsesia et al. (19).

If the CBPs are positioned in an overlapping zone of wild-type and non-wild-type populations, categorization errors will inevitably occur (14). To ensure the correct assignment of isolates in the overlapping zone, either additional investigations on the presence of potential resistance mechanisms or a relatively wide intermediate zone is required (12). Increasing knowledge of the different molecular resistance mechanisms present in clinical isolates will allow establishment of resistotype-associated RCOFF values, facilitating the correlation of the in vitro AST data with the clinical outcome, a prerequisite for the setting and validation of CBPs (31–35).