The Widely Used ATB FUNGUS 3 Automated Readings in China and Its Misleading High MICs of Candida spp. to Azoles: Challenges for Developing Countries' Clinical Microbiology Labs

The rapid development in the clinical microbiology diagnostic assays presents more challenges for developing countries than for the developed world, especially in the area of test validation before the introduction of new tests. Here we report on the misleading high MICs of Candida spp. to azoles using the ATB FUNGUS 3 (bioMérieux, La Balme-les Grottes, France) with automated readings in China to highlight the dangers of introducing a diagnostic assay without validation. ATB FUNGUS 3 is the most commonly used commercial antifungal susceptibility testing method in China. An in-depth analysis of data showed higher levels of resistance to azoles when ATB FUNGUS 3 strips were read automatically than when read visually. Based on this finding, the performance of ATB FUNGUS 3, read both visually and automatically, was evaluated by testing 218 isolates of five clinically important Candida species, using broth microdilution (BMD) following CLSI M27-A3 as the gold-standard. The overall essential agreement (EA) between ATB visual readings and BMD was 99.1%. In contrast, the ATB automated readings showed higher discrepancies with BMD, with overall EA of 86.2%, and specifically lower EA was observed for fluconazole (80.7%), voriconazole (77.5%), and itraconazole (73.4%), which was most likely due to the trailing effect of azoles. The major errors in azole drug susceptibilities by ATB automated readings is a concern in China that can result in misleading clinical antifungal drug selection and pseudo high rates of antifungal resistance. Therefore, the ATB visual reading is generally recommended. In the meantime, we propose a practical algorithm to be followed for ATB FUNGUS 3 antifungal susceptibility for Candida spp. before the improvement in the automated reading system.

antifungal resistance. Therefore, the ATB visual reading is generally recommended. In the meantime, we propose a practical algorithm to be followed for ATB FUNGUS 3 antifungal susceptibility for Candida spp. before the improvement in the automated reading system.

Introduction
The rapid development in clinical microbiology laboratory diagnostic assays presents more challenges for developing countries than their developed world counterparts, especially with regard to validation and quality control of such assays. The widely used but problematic ATB FUNGUS 3 with ATB Expression Bacteriology Analyzer automated readings (bioMérieux, La Balme-les Grottes, France) in China and its misleading reported high MICs of Candida spp. to azoles, gave a very good show case to highlight the challenges faced by clinical microbiology labs in developing countries.
Invasive candidiasis is now widely recognized as an important public health problem, with considerable morbidity, mortality, and associated health care costs [1][2][3]. Antifungal surveillance programs like ARTEMIS Global Antifungal Surveillance Program play an important role in the management of fungal infections, guiding clinical treatment as well as tracking the development of drug resistance [4][5].
Unlike in the developed world, multicenter and long term antifungal surveillance data is lacking in China. Furthermore, an analysis of reported data on antifungal susceptibility patterns from different centers in China revealed considerable variability, especially for azole drugs. ATB FUNGUS 3 is the most commonly used commercialized antifungal susceptibility test method in China, and an in-depth analysis of data on Candida susceptibility to fluconazole showed higher levels of resistance when ATB FUNGUS 3 strips were read automatically than when read visually, suggesting possible errors by the ATB FUNGUS 3 strips automated reading system [6][7][8][9][10].
Thus, the main aim of the present study was to evaluate the performance of the ATB FUNGUS 3 strips (both read visually and automatically) in relation to the Clinical and Laboratory Standard Institute (CLSI) broth microdilution (BMD) method for the in vitro antifungal susceptibility testing of Candida isolates from multicenters in China. A secondary aim was to develop a practical and efficient algorithm for Candida species susceptibility testing by ATB FUNGUS 3 strips, using a combination of visual and automated readings.

Candida isolates
A total of 218 isolates of five clinically important Candida species selected from the 2010 CHIF-NET program from 12 study centers (August 2009-July 2010) were used [11]. All the resistant isolates of the five Candida species were included in the study, whilst the susceptible isolates were selected randomly. These isolates included 81 isolates of Candida albicans, 42 of Candida glabrata, 41 of Candida tropicalis, 38 of Candida parapsilosis complex (20 of C. parapsilosis sensu stricto, 14 of Candida metapsilosis and 4 of Candida orthopsilosis) and 16 of Candida krusei. All the isolates were characterized by sequencing of the internal transcribed spacer (ITS) region.

ATB FUNGUS 3
An ATB FUNGUS 3 strip consists of 16 pairs of cupules including two growth control wells and five antifungal drugs at different concentrations: 5-flucytosine (4, 16 mg/mL), amphotericin B (0.5-16 mg/mL), fluconazole (1-128 mg/mL), itraconazole (0.125-4 mg/mL) and voriconazole (0.06-8 mg/mL). Testing on the ATB FUNGUS 3 strip was performed simultaneously with CLSI BMD. Following the manufacturer's instructions, a suspension with a turbidity of 2 McFarland was prepared and 20 mL of this suspension was transferred to an ampule of ATB FUNGUS 3 Medium. After this, 135 mL of the inoculated medium was transferred into each cupule.
After incubation at 35˚C for 24 h, the strips were read both visually and automatically on the ATB Expression Bacteriology Analyzer automatic system (bioMérieux, La Balme-les Grottes, France). According to the manufacturer's instructions, the MICs were determined by the growth score for each of the cupules compared with the control cupules ( Figure 1). For amphotericin B, the MIC corresponds to the lowest concentration enabling complete growth inhibition (score5''0''). For 5-flucytosine, fluconazole, itraconazole and voriconazole, the MIC corresponds to the lowest concentration of the antifungal agent with which a score of ''2'', ''1'', ''0'' is obtained.

Quality control
Quality control was ensured by testing the CLSI-recommended quality control strains C. parapsilosis ATCC 22019 and C. krusei ATCC 6528 for both CLSI BMD and ATB FUNGUS 3 [12].

Data analysis
The results obtained with the ATB FUNGUS 3 (both visual and automated readings) were compared with those of CLSI BMD after 24 h incubation. The results were considered to be in essential agreement (EA) when the ATB FUNGUS 3 results were within ¡2 dilutions of the BMD reference value (eg. visual reading51 mg/mL, and automated reading54 mg/mL, defined as in EA; whilst visual reading51 mg/mL, automated reading58 mg/mL, defined as not in EA). The rates of EA (%) were the percentage of the isolates in EA with BMD. The results were considered to be in categorical agreement (CA) when the ATB FUNGUS 3 and CLSI BMD results fell within the same interpretive category (i.e., susceptible, susceptible dose dependent [SDD], intermediate, or resistant, depending on the drugs tested). If established CLSI M27-S4 interpretive breakpoints were available, they were used to determine CA. Epidemiological cutoff values (ECVs) were used for species/drugs without CLSI breakpoints (the susceptibility results are defined as either wide type [WT] or non-WT by ECVs) [13,14]. Very major errors were identified when BMD indicated a resistant/non-WT result and the ATB FUNGUS 3 indicated a susceptible/WT one. Major errors were identified when BMD showed a susceptible/WT result while the ATB FUNGUS 3 showed the opposite. Minor errors were identified when one method recorded a susceptible or resistant result whilst the other recorded a SDD result.

Proposing a Candida species susceptibility testing algorithm
We developed an algorithm based on combination of ATB FUNGUS 3 automated and visual readings, which could be used when the isolates are identified to species level ( Figure 2). Using this algorithm, the results are initially read by ATB automated reading system, and the isolates were put into categories [13,14]. The isolates with non-resistant/WT results for all drugs can be reported as the final results, but those which show resistant results for any drug should be checked by visual reading. If there is still difficulty in the visual reading or the resistance type is atypical, then the susceptibility result should be confirmed by BMD assay.

Data reviewed
As shown in Table 1, using the ATB automated reading system, the reported level of resistance to fluconazole reached as high as 42.3% for C. albicans and 52.6% for C. tropicalis, which is in sharp contrast to the results from visual readings (highest: 6.6% for C. albicans and 10.2% for C. tropicalis). Furthermore, published surveillance data for 12 institutions in China (CHIF-NET) showed fluconazole resistance rates of 0.4% (C. albicans) and 5.7% (C. tropicalis). The ARTEMIS global antifungal surveillance program also showed fluconazole resistance rates of only 1.4% for C. albicans and 4.1% for C. tropicalis (Table 1).

Essential agreement (EA) between ATB FUNGUS 3 and CLSI BMD
The overall EA between ATB visual readings and BMD was 99.1%. The highest concordance between the two methods for all the Candida species tested was observed for amphotericin B (100%) and 5-flucytosine (100%) ( Table 2). Of all the Candida species tested, C. glabrata showed the lowest concordance between the two methods ( Table 2 and 3). The overall EA between ATB automated readings and BMD was only 86.2%, which was significantly lower (P,0.05) than those of visual readings (99.1%). The highest agreement was observed for amphotericin B (99.5%) and 5-flucytosine (99.5%), but much lower EA for fluconazole (80.7%), voriconazole (77.5%), and itraconazole (73.4%). Most isolates with discrepant results had MICs of more than 2 dilutions higher than that of BMD for azole drugs. C. tropicalis isolates showed the lowest EA, with only 31.7% for fluconazole, 24.4% for voriconazole and 22.0% for itraconazole. In contrast, C. parapsilosis and C. krusei showed very high EA levels with BMD (Tables 2 and 3).

Categorical agreement (CA) between ATB FUNGUS 3 and CLSI BMD
The overall CA for the comparisons of the ATB visual reading results with the BMD results was 97.5%, and 95.4% and 95.0% for fluconazole and voriconazole, respectively ( Table 2). In contrast, the overall CA between the ATB automated readings and the BMD results was only 84.0%, and 78.4% and 72.9% for fluconazole and voriconazole, respectively. The highest major error rate was observed for susceptibility of C. tropicalis to either fluconazole (23 of 41 isolates studied, 56.1%) or voriconazole (24 of 41, 58.5%), followed by C. albicans (

Validation of the proposed algorithm
Based on this algorithm, all C. tropicalis isolates can be directly read visually because of poor accuracy by the ATB automated reading system. However, the local prevalence of C. tropicalis should be considered when choosing this proposal. When we applied the proposed algorithm to the 218 isolates in the present study Table 1. Reported susceptibilities of C. albicans and C. tropicalis to fluconazole using ATB visual and automated readings in single centers of China and susceptibilities by disc diffusion in global and national antifungal surveillance programs.

Method
Study site Susceptibility to fluconazole (% resistant) Reference    Table 4, the algorithm offered a very good balance between accuracy and automation. As can be seen, the proposed testing algorithm is practical, labor and time saving for Candida susceptibility testing in large surveillance programs (Table 4).

Discussion
ATB FUNGUS 3 is the most commonly used commercialized antifungal susceptibility test method in China, and is also widely used in parts of Africa, South America and Europe [16][17][18][19]. A former version (ATB FUNGUS 2) of this current system has been evaluated by Torres-Rodriguez et al. [17]. In agreement to our findings, lower agreement with BMD was observed for fluconazole and itraconazole, particularly with C. tropicalis and C. albicans. [17].  In the present evaluation, ATB visual readings showed good concordance with BMD, whilst a high pseudo resistant rate (defined as major errors) to azole drugs was observed for ATB automated readings. The potential explanation for the high major errors with the ATB automated readings is the trailing effect of azoles. We found that C. tropicalis and C. albicans showed obvious trailing growth in testing with azole drugs. The activation of calcineurin and altered regulation of genes mediating resistance could partially explain the trailing phenomenon [20,21].
The major errors in azole drug susceptibility testing by ATB automated readings are a cause for concern for several reasons. Firstly, the very high pseudoresistance may result in misleading clinical antifungal drug selection and increased costs for antifungal drugs. Secondly, national and international surveillance studies and epidemiological comparison studies may potentially get misleading results and conclusions if they simply used the ATB automated reading datagenerally will over-estimate the China azoles antifungal resistance rates. Thirdly, the unwarranted use of alternative antifungal agents may promote further resistance in these agents, making future drug selection difficult [22]. Therefore, the ATB visual reading is generally recommended.
The present study highlights the flaws encountered in commercial automated identification and susceptibility testing systems especially in developing countries. Very recently, similar problems were reported by Chowdhary et al. in Asia, where VITEK 2 (bioMérieux, Marcy I'Etoile, France) misidentified 10 Candida auris isolates as Candida haemulonii and 2 as Candida famata [23,24]. The problems highlighted the need for an improvement and validation of the automated systems as well as alerting laboratory staff of the potential errors. Before the improvement of the automated reading system, we developed an algorithm to make good use of the ATB reading system, which is especially practical for surveillance programs.
Generally, the ATB FUNGUS 3 (visual reading) is a simple and accurate method for the determination of MICs for Candida spp., which is suitable for clinical microbiology use in developing countries like in Asia, Africa and South America. The present study served as an important experience in ATB FUNGUS 3 practical use for other countries where it is routinely used.