INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

With the recent global resurgence of tuberculosis (TB) and concomitant rise in multidrug-resistant strains of Mycobacterium tuberculosis (37, 38), there is an increasing demand for determining the in vitro susceptibilities of clinical isolates to antimicrobial agents other than those considered primary drugs (isoniazid, rifampin, ethambutol, pyrazinamide, and streptomycin). Conventional antimicrobial susceptibility testing (AST) with solid media such as Löwenstein-Jensen (LJ) or Middlebrook 7H10 agar requires 3 or more weeks to be completed, and for some classical second-line drugs as well as newer antimicrobial agents, appropriate critical drug concentrations have not fully been established.

AST by the radiometric BACTEC method (Becton Dickinson Diagnostic Instrument Systems, Sparks, Md.), introduced in 1980, is an efficient way to test frontline drugs (27, 30-33) and results in a significantly shorter turnaround time than that of the traditional proportion method. Susceptibility testing by this procedure has become the current method of choice (34), even when the hazards arising from the use of radioisotopes are taken into account. However, data on critical concentrations for classical second-line and other newer drugs are largely fragmentary or lacking altogether. Some data from AST of second-line drugs with BACTEC 12B medium were presented quite early (11) but have never been published. Allen et al. (1) defined the MICs on Middlebrook agar of aminoglycosides while Heifets' group undertook several investigations into AST with the BACTEC 12B medium, e.g., with aminoglycosides (13), ethionamide (16, 22), and quinolones (15), the last group of compounds having also been studied by others (4, 5, 8, 9). The MICs of rifabutin, generated either with the radiometric BACTEC 12B medium (17) or with Middlebrook 7H9 broth (7), are also available. However, as a whole, most of these studies have either included only a narrow spectrum of drugs and/or tested only a rather limited number of M. tuberculosis strains.

In light of this situation, the primary aim of our study was to develop a basic protocol which defines appropriate critical concentrations for secondary drugs to allow reliable testing with both BACTEC 12B and agar medium. With more than 270 strains of M. tuberculosis, the reproducibility of results generated by different laboratories was also assessed. Once validated, the radiometric procedure should allow clinical laboratories to provide physicians with accurate and timely drug susceptibility information. In addition, since quite a few nonradiometric mycobacterial culture systems based on liquid medium, such as the Mycobacteria Growth Indicator Tube (manual version [23]; BACTEC 960 automated version [12]), the MB/BacT (2), and the ESP Culture System II (35), have recently been developed, this study may also provide a guideline for future development of AST procedures with these novel devices.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Study sites. This study was conducted at six mycobacteriology laboratories, including the Mycobacteriology Laboratory, Veterans Affairs Medical Center, West Haven, Conn.; the Bureau of Laboratories, New York City Department of Health, New York, N.Y.; the Laboratory Center for Disease Control, Ottawa, Ontario, Canada; the Mycobacteriology Laboratory, Mayo Clinic, Rochester, Minn.; the Mycobacteriology Laboratory, Wadsworth Center, New York State Department of Health, Albany, N.Y.; and the Swiss National Center for Mycobacteria, Department of Medical Microbiology, University of Zurich, Zurich, Switzerland. A seventh site, the Research and Development Division, Becton Dickinson Microbiology Systems, Sparks, Md., was the coordination site, from which cultures and antimicrobial agents were supplied and at which data were collected.

Antimicrobial agents. All drugs were obtained from the manufacturers in a chemically pure form. The drugs used were capreomycin sulfate, D-cycloserine, ethionamide, and kanamycin monosulfate from Sigma, St. Louis, Mo.; amikacin from Bristol-Myers Squibb, Syracuse, N.Y.; clofazimine from Ciba-Geigy, Suffern, N.Y.; ciprofloxacin from Miles, West Haven, Conn.; ofloxacin from Ortho/R. W. Johnson Pharmaceutical, Raritan, N.J.; and rifabutin from Adria Laboratories, Dublin, Ohio. The compounds were supplied by the coordinator of the study. Amikacin, capreomycin, ciprofloxacin, and kanamycin were dissolved in sterile distilled water (DW). Ofloxacin was dissolved in a 0.1 N NaOH solution. Ethionamide was dissolved in ethylene glycol and incubated overnight for sterilization. Subsequent dilutions of these two drugs were made in sterile DW. Clofazimine was dissolved in dimethyl sulfoxide and stored at room temperature in the dark. Subsequent dilutions were made in the same solvent. For D-cycloserine a 0.1% solution of Na₂CO₃ was prepared and added to 100 ml of DW until a pH of 10 was achieved. Dilutions were made in sterile DW. Rifabutin was dissolved in methanol and diluted in DW. Stock solutions of each drug were made at least 40 times higher than the highest test concentration used. Except for clofazimine and ethionamide, which are considered self-sterilizing, all stock solutions were filtered in a sterile manner with a 0.22-µm-pore-size polycarbonate filter; approximately 20% of the initial filtrate was discarded. All stock solutions except clofazimine were stored at 70°C in small aliquots. Once thawed, the remaining solutions were discarded.

Culture media and quality control. Middlebrook 7H12 broth (BACTEC 12B; Becton Dickinson Microbiology Systems) was used for radiometric testing, and Middlebrook 7H10 agar plates (BBL, Cockeysville, Md.) were used for AST by the proportion method. All media were supplied by the coordinator of the study. M. tuberculosis H37Rv (ATCC 27294; susceptible to all drugs tested) was used as a strain for internal quality control, which was done on a weekly basis, along with the other strains included in the study.

AST. (i) BACTEC 460. For radiometric AST, the standard protocol was followed (31). Each drug was tested at several concentrations. Antimicrobials were always added in 0.1-ml quantities to a BACTEC 12B vial to achieve the desired concentrations. Susceptibility and resistance were judged by comparison of the change in the growth index of the control with that of the test drug as recommended. This interpretation gave susceptibility results at the 1% proportion basis.

(ii) Traditional proportion method. AST was carried out with Middlebrook 7H10 agar according to the standard procedure (18, 20). For some classical second-line antituberculotics (capreomycin, cycloserine, ethionamide, and kanamycin), the recommended critical concentrations (20) were used. For all other drugs, several concentrations were tested to establish the MICs and critical concentrations. The MIC was defined as the lowest concentration of drug that inhibited more than 99% of the bacterial population. As a consequence, the critical concentration was defined as the concentration of drug required to prevent growth above the 1% threshold (critical proportion) of the test population of TB bacilli (18). Results were read 21 days after inoculation of the medium.

Study design and strains. The study was carried out concomitantly in three phases at all six centers.

Phase I. Phase I was designed to establish a basic test procedure and to determine a working range for the antimicrobial drug concentrations to be used in liquid medium (BACTEC 12B) and on conventional solid media. In addition, the reproducibility of the results of the test systems was evaluated. For this purpose, a total of 10 clinical isolates of M. tuberculosis which were susceptible to all primary drugs in earlier tests and stored in a strain collection were subcultured onto LJ slants. As soon as growth appeared on the slants (10 to 15 days after inoculation) these isolates were shipped by overnight delivery to the six clinical testing sites. In an effort to minimize variations of results, the cultures were used directly for AST without subculturing. For those antibiotics whose critical concentrations had been established previously on solid medium, i.e., capreomycin, cycloserine, ethionamide, and kanamycin, only one drug concentration was tested on Middlebrook 7H10 agar but at least three concentrations were tested in BACTEC 460. For all other drugs at least three different concentrations were tested by the proportion method as well as with BACTEC 460.

Phase II. The aims of phase II were to establish equivalent breakpoint drug concentrations for the BACTEC 460 and conventional solid medium methodologies and to evaluate interlaboratory reproducibility of test results for susceptible and resistant isolates. Drug concentrations used in phase II were adjusted and finalized based on the information obtained from phase I. Strains of M. tuberculosis with known drug resistance, especially to the test drugs, were selected from several clinical sources. Twenty isolates were selected and subcultured onto LJ medium. Once growth appeared, sets of 20 isolates each were shipped to the six test sites and were used for AST as in phase I.

Phase III. Phase III was extended testing with both BACTEC 460 and Middlebrook 7H10 agar with the drug concentrations established in phase II. Staff at each site independently tested at least 24 strains of M. tuberculosis which had recently been isolated from clinical specimens at their own laboratory.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

In phases I and II of this multicenter AST study sets of TB strains containing 10 and 20 isolates, all from stock cultures, were analyzed. Strains tested in phase I were confirmed to be susceptible to all drugs. Among the 20 strains tested in phase II, 6 were resistant to capreomycin, 0 were resistant to cycloserine, 10 were resistant to ethionamide, 9 were resistant to kanamycin, 8 were resistant to amikacin, 1 was resistant to clofazimine, 1 was resistant to ofloxacin, and 13 were resistant to rifabutin. In phase III, AST was performed with a total of 242 M. tuberculosis strains which had recently been isolated from clinical specimens by staff at the six participating laboratories. Nineteen strains were resistant to capreomycin, 19 were resistant to cycloserine, 56 were resistant to ethionamide, 34 were resistant to kanamycin, 32 were resistant to amikacin, 0 were resistant to clofazimine, 20 were resistant to ofloxacin, and 83 were resistant to rifabutin.

Phase I. The 10 fully drug-susceptible isolates supplied to all sites were tested against various concentrations of eight drugs. Most BACTEC 460 MICs determined at the test sites agreed within ±1 serial dilution (Table 1). For example, the MIC of amikacin was 1.0 µg/ml for 100% of the strains and that of capreomycin was 1.25 µg/ml for 98.3% of the strains at all test sites combined. With a few drugs, MICs extended over a wider range than that noted above. This held true for the quinolones, whose MICs varied by more than 1 dilution from site to site, although in the majority of cases the values were between 0.5 and 1.0 µg/ml. Cycloserine gave the least-reproducible results for both BACTEC 460 and conventional testing (proportion method). Based on phase I results, drug concentrations were adjusted for subsequent testing in phase II. For amikacin, clofazimine, and ethionamide, drug concentrations to be used with BACTEC 460 were lowered by 1 dilution, whereas for cycloserine and ofloxacin, concentrations were increased for both solid and liquid medium systems. Ciprofloxacin testing was discontinued; on the other hand, rifabutin testing was included in phases II and III.

Phase II. Results of phase II testing of the 20 susceptible and resistant strains of M. tuberculosis sent to all six study sites are reported in Table 2. Interlaboratory reproducibility of results by the BACTEC 460 and conventional AST methods was good, with MICs being within ±1 dilution of each other in most cases. It appeared that agar AST results varied more from site to site than those obtained by the BACTEC 460 method. In general, interlaboratory variation in results was observed more frequently with cycloserine and to some extent with ethionamide and capreomycin than with the other drugs, regardless of the AST method applied.

Phase III. Phase III extended testing with 242 clinical isolates of M. tuberculosis among the six study sites, with 27 isolates at site 1, 42 isolates at site 2, 48 isolates at site 3, 24 isolates at site 4, 68 isolates at site 5, and 33 isolates at site 6. Tentative breakpoint concentrations elaborated in phase II (Tables 3 and 4) were acceptable with some minor changes (Table 5). Generally, we looked at the concentrations which gave the fewest total errors (ethionamide at 1.25 µg/ml in the BACTEC 460 and 5.0 µg/ml in conventional AST, kanamycin at 5.0 µg/ml in the BACTEC 460 and 5.0 µg/ml in conventional AST, amikacin at 1.0 µg/ml in the BACTEC 460 and 4.0 µg/ml in conventional AST, clofazimine at 0.5 µg/ml in the BACTEC 460 and 1.0 µg/ml in conventional AST, ofloxacin at 2.0 µg/ml in the BACTEC 460 and 2.0 µg/ml in conventional AST, and rifabutin at 0.5 µg/ml in the BACTEC 460 and 1.0 µg/ml in conventional AST). For kanamycin, amikacin, clofazimine, ciprofloxacin, and rifabutin these concentrations also gave the fewest very major errors. The only exception was capreomycin. Although overall agreement for this drug at 2.5 µg/ml (8 very major errors, 2 major errors) was better than at 1.25 µg/ml (4 very major errors, 13 major errors), we have chosen 1.25 µg/ml as the breakpoint (only 4 very major errors instead of 8). Cycloserine testing was, again, very inconsistent by both the BACTEC 460 and agar methodologies. At one site, cycloserine testing by the method of proportion did not yield reportable results at all, with false resistance being shown even with the M. tuberculosis H37Rv control. These data were excluded from our analysis (Table 5).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The data presented here constitute the first comprehensive, multicenter AST study of second-line anti-TB drugs and newer compounds currently being used for the treatment of TB, validating both broth-based and solid-medium-based systems. Given the fact that MICs, as a whole, are highly dependent on a large array of different factors such as medium composition, pH, inoculum size, and incubation time, the primary aim of this study was to establish critical concentrations with which to perform AST on a wide spectrum of anti-TB drugs under defined conditions.

Our data, encompassing more than 17,000 individual susceptibility test results generated with 272 strains of M. tuberculosis in six centers, corroborate the fact that the radiometric BACTEC 460 method is feasible for AST of most of these drugs as it has previously been shown to be for frontline anti-TB drugs (30-33). Complying with the recommendation of the Centers for Disease Control and Prevention of using broth-based methods for AST (34), more and more laboratories in the United States have now adopted the BACTEC 460 technique so that test results can be reported to the clinician with a minimal turnaround time, i.e., within 4 to 8 days, as was the average in our study. Besides the longer turnaround time, there are several other drawbacks of AST on solid media, such as drug inactivation during agar preparation and degradation of the antimicrobial agent during the extended period of incubation (10), which can, at least to some extent, be circumvented easily by using broth-based media. Another important goal of the study was to achieve a high degree of interlaboratory reproducibility of AST results. As indicated by our data, the agreement in results among the six laboratories was quite high throughout the three phases.

The drug panel remained the same throughout the study, except that (i) ciprofloxacin was dropped after phase I because of its known poor in vivo response against TB, and (ii) rifabutin, a promising anti-TB drug, was added for phases II and III. For the first two phases, each of the six sites was provided with identical sets of M. tuberculosis strains. In testing of these strains by all participants, susceptibility data generated in these two initial phases were found to be highly reliable and reproducible. Further testing eventually led to the establishment of the final critical test concentrations for seven of the eight second-line drugs. Basically, establishment of the final concentrations was achieved in three major steps: (i) by establishing MICs by testing fully susceptible TB strains (phase I), (ii) by adjusting drug concentrations and testing susceptible and resistant strains (phase II), and (iii) by testing a large number of clinical M. tuberculosis isolates (n = 242) with particular emphasis on isolates that are resistant to primary drugs. This process helped in establishing the final breakpoints by selecting those drug concentrations which gave minimum false susceptible and false resistant results with both test systems.

Allen et al. (1) have defined critical concentrations in Middlebrook 7H11 agar as 9.5 µg/ml for capreomycin, 2.5 µg/ml for kanamycin, and 1.0 µg/ml for amikacin. These concentrations are different from ours on Middlebrook 7H10, as we tested 10.0 µg of capreomycin per ml and 5.0 µg of kanamycin per ml, which have been established previously as critical concentrations, and 4.0 µg of amikacin per ml.

Important work in radiometric AST, though with a very limited number of strains of M. tuberculosis, has been carried out by Heifets et al. (13-17). This group has suggested the following critical drug concentrations for BACTEC 12B medium: 5.0 µg/ml (MIC range, 1.5 to 3.0) for capreomycin, 4.0 µg/ml (indicating an MIC range of 0.5 to 1.0) for amikacin, 2.0 µg/ml (MIC range, 0.25 to 2.0) for ofloxacin, and 0.12 µg/ml (MIC range, 0.015 to 0.06) for rifabutin. In general, these MICs differ by 1 to 2 serial dilutions from ours. This divergence is not surprising when one considers those authors' notion that, at these suggested critical concentrations, the test strain, if susceptible, should be considered only moderately susceptible. In their hands, fully susceptible strains were those which were susceptible to a concentration 1 dilution (twofold) lower than the critical concentration while resistant strains were those which were resistant to a concentration 1 serial dilution higher than the critical concentration. Even a fourth category, designated very resistant, was suggested. Though the concept of four categories of susceptibility, i.e., susceptible, moderately susceptible, resistant, and very resistant, might be helpful to some physicians, in particular when dealing with multidrug-resistant TB, this concept has, however, not been widely accepted.

Additional information is available for the quinolones (4, 5, 19, 21). Chen et al. (5) established MIC ranges in BACTEC 12B medium for ciprofloxacin between 0.25 and 2.0 µg/ml and for ofloxacin between 0.5 and 2.0 µg/ml and set the breakpoint concentration at 2.0 µg/ml. These values agree perfectly with our ofloxacin data (MICs, 1.0 to 2.0 µg/ml; breakpoint, 2.0 µg/ml). Others, however, have reported MICs of ofloxacin between 0.5 and 1.0 µg/ml and a critical concentration of 1.0 µg/ml (4). The range of MICs for our test strains on solid medium appeared, however, to be narrow (1.0 to 4.0 µg/ml, with a critical concentration of 2.0 µg/ml), compared to MICs of 0.3 to 4.0 µg/ml and 0.125 to 4.0 µg/ml for ciprofloxacin and ofloxacin, respectively, in Middlebrook 7H10 medium (5).

The reported MICs of rifabutin in BACTEC 12B medium extend over a broad range. Our values (0.5 to 1.0 µg/ml) are perfectly concordant with those of Della Bruna and Olliaro (6). However, other groups have found both lower ranges (0.015 to 0.125 µg/ml [5] and 0.015 to 0.6 µg/ml [17]) and a higher range (3.6 µg/ml [7]), the last having been generated in Middlebrook 7H9 broth. One of the most recent studies of AST of secondary anti-TB drugs (25) confirms largely our values for capreomycin (1.0 to 2.0 versus 1.25 to 5.0 µg/ml [this study]), kanamycin (5.0 versus 2.0 to 4.0 µg/ml), amikacin (0.5 to 1.0 versus 0.5 µg/ml), and clofazimine (0.5 versus 0.1 to 0.4 µg/ml). Such low MICs of the latter drug were also found by others (26). Rastogi et al. (25) used a human macrophage system and demonstrated that, of a large variety of drugs tested, clofazimine was the only one which yielded discrepant results between its extracellular and intracellular activities. This may be due to the fact that this drug concentrates within phagosomes and phagolysosomes of infected macrophages (24), which might, at least to some extent, explain its ineffectiveness as an anti-TB agent.

Even though for some of the compounds tested it was not possible to include a fair amount of resistant strains in the test panels, which would call for additional studies with genetically well-characterized, resistant strains, several important conclusions can be drawn from our study. (i) Except with cycloserine, the radiometric BACTEC 460 procedure is a simple and rapid method requiring 4 to 8 days on average to generate reliable AST results for second-line anti-TB drugs. (ii) Cycloserine testing should be discontinued, due to difficulties with the drug at all six study sites, despite repeated testing. (iii) With these comprehensive AST data it is anticipated that our evaluation may also provide a guideline for future studies, especially when AST procedures are established for the newer, nonradiometric, broth-based systems such as the MB/BacT (28), the ESP Culture System II (3), the manual Mycobacteria Growth Indicator Tube (29, 36) or its automated version (BACTEC 960). The drug concentrations defined here for BACTEC 12B medium and Middlebrook 7H10 agar will undoubtedly serve as a reliable baseline for establishing critical test concentrations for AST by these recently developed growth-based technologies.