|Home | About | Journals | Submit | Contact Us | Français|
The aminoglycosides and cyclic polypeptides are essential drugs in the treatment of multidrug-resistant tuberculosis, underscoring the need for accurate and reproducible drug susceptibility testing (DST). The epidemiological cutoff value (ECOFF) separating wild-type susceptible strains from non-wild-type strains is an important but rarely used tool for indicating susceptibility breakpoints against Mycobacterium tuberculosis. In this study, we established wild-type MIC distributions on Middlebrook 7H10 medium for amikacin, kanamycin, streptomycin, capreomycin, and viomycin using 90 consecutive clinical isolates and 21 resistant strains. Overall, the MIC variation between and within runs did not exceed ±1 MIC dilution step, and validation of MIC values in Bactec 960 MGIT demonstrated good agreement. Tentative ECOFFs defining the wild type were established for all investigated drugs, including amikacin and viomycin, which currently lack susceptibility breakpoints for 7H10. Five out of seven amikacin- and kanamycin-resistant isolates were classified as susceptible to capreomycin according to the current critical concentration (10 mg/liter) but were non-wild type according to the ECOFF (4 mg/liter), suggesting that the critical concentration may be too high. All amikacin- and kanamycin-resistant isolates were clearly below the ECOFF for viomycin, and two of them were below the ECOFF for streptomycin, indicating that these two drugs may be considered for treatment of amikacin-resistant strains. Pharmacodynamic indices (peak serum concentration [Cmax]/MIC) were more favorable for amikacin and viomycin compared to kanamycin and capreomycin. In conclusion, our data emphasize the importance of establishing wild-type MIC distributions for improving the quality of drug susceptibility testing against Mycobacterium tuberculosis.
The emergence of multidrug-resistant (MDR) tuberculosis (TB) and extensively drug-resistant (XDR) Mycobacterium tuberculosis creates higher demands for accurate and reproducible antimicrobial drug susceptibility testing (DST) (11, 12, 25). Injectable drugs, such as aminoglycosides and cyclic polypeptide antibiotics, are, together with fluoroquinolones, the most important class of second-line drugs in the treatment of MDR TB and are needed to define XDR TB. If possible, injectable drugs are used in combination with at least three active oral drugs for at least 18 months to treat MDR TB (2, 7). The mechanism of action for aminoglycosides and cyclic polypeptides is mainly the inhibition of protein synthesis (1). As a class member of the aminoglycosides, streptomycin was the first drug used for the treatment of tuberculosis in the 1940s (24). Until the recommendations were revised by the WHO, streptomycin was interchangeable with ethambutol as a first-line drug but was abandoned due to increasing resistance and the risks involved in administering injections in areas where HIV is endemic (7). Where afforded, amikacin is commonly the injectable drug of choice in the treatment of MDR TB. According to WHO recommendations (25), amikacin lacks a critical concentration for Middlebrook 7H10 medium, and no breakpoint in any DST method is available for viomycin (20, 25). Resistance to the aminoglycosides and cyclic polypeptides is associated with mutations in the 16S rRNA gene (rrs), and, in addition, mutations in the tlyA gene also confer resistance to the cyclic polypeptides (8, 15). It has been suggested that isolates with low-level and intermediate-level kanamycin resistance can be susceptible to amikacin, whereas strains with high-level kanamycin resistance are resistant to both drugs (13, 15). Cross-resistance among the cyclic polypeptides has been reported between capreomycin and viomycin, but limited data are available (16).
DST against M. tuberculosis is currently based on testing the susceptibility to single so-called critical concentrations of antibiotics (17). However, the scientific basis for defining these critical concentrations is weak, particularly for second-line drugs (25). Using inappropriate susceptibility breakpoints might lead to poor reproducibility and the incorrect reporting of susceptibility results to clinicians (25). In a first attempt to address this issue, we recently reported the problematic close relationship between the wild-type MIC distribution and the current critical concentration for ethambutol (23). In the determination of clinical breakpoints for DST with regard to most other bacterial pathogens, wild-type MIC distributions represent a significant and necessary tool (9). The definition of a wild-type strain is a microorganism without acquired and mutational resistance mechanisms to a certain drug. The wild-type cutoff, which is commonly labeled the epidemiological cutoff value (ECOFF), is one important tool in addition to pharmacokinetic and pharmacodynamic (PK/PD), as well as clinical, data used by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) when setting clinical breakpoints (9). Surprisingly, published wild-type MIC distributions for M. tuberculosis are limited, but the lack of such data is commonly discussed (4, 11, 21, 25). Clinical outcome data in relation to MIC determinations are very difficult to achieve for second-line drugs against MDR TB because several drugs are used in combination. As a result, wild-type MIC distributions and PK/PD data are essential in order both to establish reasonable breakpoints for M. tuberculosis and to be able to predict susceptibility patterns for novel anti-TB drugs. Such data could shed some light on the conflicting reports of cross-resistance within the aminoglycoside group, for which the lack of background data in establishing the current critical concentrations could be a significant cause of the variable results (8, 15, 16).
In the present study we aimed to evaluate the current critical concentrations for injectable aminoglycosides and cyclic polypeptides used in the treatment of M. tuberculosis by comparing them to wild-type MIC distributions and available PK/PD data.
To establish the wild-type distribution, 95 consecutive clinical isolates from the clinical mycobacteriology laboratory at Karolinska University Hospital were included. Five isolates were excluded due to clustering according to the restricted fragment length polymorphism (RFLP) database at the Swedish Institute for Disease Control as previously described (23). Additionally, 21 poly- and multidrug-resistant strains were included. The MIC determinations were performed in two separate runs after blinded renumbering. In each run, the pan-susceptible M. tuberculosis H37Rv strain (ATCC 27249) was tested in duplicate. In the second run, six strains were tested in duplicate along with three consecutive strains from the same patient.
Susceptibility testing to first-line drugs was performed previously using the Bactec 460 method (B460) (Becton Dickinson). The susceptibility to amikacin (1.0 mg/liter) was tested on isolates that were resistant to any first-line drug, while the susceptibility to capreomycin (1.25 mg/liter) and streptomycin (2 mg/liter) was tested on MDR strains.
Stock solutions (30.7 g/liter) in distilled water were prepared for amikacin, kanamycin, capreomycin, streptomycin (all from Sigma, St. Louis, MO), and viomycin (VWR International, Stockholm, Sweden). The methodology for MIC determination is described in detail elsewhere (23). Briefly, bacterial suspensions of all strains were transferred by using a 96-stick replicator to Middlebrook 7H10 agar plates with 1:2 serial dilutions from 0.002 to 512 mg/liter for all drugs. Drug-free control plates with undiluted and 1:100 diluted bacterial suspensions were replicated in the same way. Inoculated agar plates were incubated at 37°C for 3 weeks. The MIC was defined as the first antibiotic concentration which showed less growth compared to the 1:100 diluted controls of the corresponding strain, i.e., the lowest concentration of drug that inhibited more than 99% of the bacterial population. MIC testing against amikacin as a class representative was performed on a subset of 10 isolates, including H37Rv in Bactec MGIT 960 (MGIT). Serial dilutions of amikacin were prepared as outlined above, and MGIT tubes were inoculated on the day of the experiment in final concentrations corresponding to two MIC steps above and below the MIC obtained in the Middlebrook 7H10. The control strain M. tuberculosis H37Rv was included in all runs as a quality control.
The peak serum concentration (Cmax) was chosen as the important PK parameter based on data availability and the fact that it is commonly used to define PD indices for other bacteria (10, 18). The Cmax divided by the MIC was calculated for amikacin, kanamycin, streptomycin, capreomycin, and viomycin according to standard doses and previously published pharmacokinetic data (3, 18, 19). The free fraction was calculated by multiplying the percentage of protein-free drug fraction with the Cmax, and the free fraction of this variable (fCmax) was divided by the corresponding MIC value (see Table Table44 for the pharmacokinetic data used for calculations). All parameters were used in the lower range of published data (3, 6, 18, 19).
Overall, the MIC variation between and within runs did not exceed one MIC dilution step (Table (Table1).1). H37Rv was tested four times, and seven isolates, including resistant and susceptible strains, were tested twice. Three consecutive isolates with aminoglycoside resistance, representing separate cultures within a year, from an MDR TB patient, showed excellent reproducibility for viomycin and capreomycin, while the MICs for the other tested drugs were above 64 mg/liter at all occasions.
For all injectable drugs investigated (streptomycin, viomycin, capreomycin, kanamycin, and amikacin), the MICs of consecutive susceptible clinical isolates formed a distinct normal distribution—the wild-type MIC distribution (Fig. 1a to e). Tentative ECOFFs separating wild-type from non-wild-type strains were defined for all investigated drugs (Table (Table2),2), including viomycin and amikacin, for which no recommended critical concentration for solid medium is available. The ECOFFs (i.e., the highest MIC of the wild-type distribution) were defined as 4 mg/liter for kanamycin and capreomycin, 2 mg/liter for streptomycin and viomycin, and 1 mg/liter for amikacin. Five out of seven amikacin- and kanamycin-resistant isolates were classified as susceptible to capreomycin according to the current critical concentration for 7H10 (10 mg/liter) but had MIC levels at 8 mg/liter, which was non-wild type according to the ECOFF (4 mg/liter). All seven amikacin- and kanamycin-resistant isolates were clearly below the ECOFF (2 mg/liter) for viomycin (Table (Table2).2). Two isolates clearly below the ECOFF (2 mg/liter), as well as below the critical concentration, for streptomycin were resistant to capreomycin, kanamycin, and amikacin but below the ECOFF for viomycin (Table (Table22).
A comparison of the results from MIC determination for amikacin in the Middlebrook 7H10 medium using the tentative ECOFF (1 mg/liter) with susceptibility results in B460 (1 mg/liter) showed full agreement (29/29, including 7 resistant isolates). This was also the case for streptomycin, for which a total agreement between Middlebrook 7H10 and B460 was observed (11/11, including 10 resistant isolates according to the ECOFF and the critical concentration). Regarding capreomycin, the agreement was 80% (8/10), for which the two discordant isolates resistant to capreomycin in B460 had MICs of 8 mg/liter in Middlebrook 7H10, which would be interpreted as susceptible according to the current critical concentration (10 mg/liter) but non-wild type according to the ECOFF for capreomycin (4 mg/liter). Furthermore, these two strains were highly resistant to kanamycin and amikacin but had MICs below ECOFF to viomycin (Table (Table22 and Fig. 1a to e). There was a good agreement of MIC determinations for amikacin between Bactec MGIT 960 and Middlebrook 7H10 for susceptible and resistant strains (n = 9; Table Table33).
The results from the PK/PD calculations are shown in Table Table4.4. It should be noted that no pharmacodynamic targets have been determined for M. tuberculosis, although a Cmax/MIC at a minimum of 8 to 10 is a generally accepted target for the aminoglycoside group (10, 18). The Cmax/MIC target for cyclic polypeptides is unknown. In general, the selected PD index (fCmax/MIC) was most favorable for amikacin, followed by streptomycin and viomycin, since a higher fCmax/MIC ratio among wild-type isolates was achieved for these drugs than for kanamycin and capreomycin (Table (Table4).4). For both kanamycin and capreomycin, not all wild-type strains were covered by the target of an fCmax/MIC value of ≥10 (Table (Table44).
In an effort to reevaluate the current critical concentrations used in drug susceptibility testing for Mycobacterium tuberculosis, we established wild-type MIC distributions of the most commonly used aminoglycosides and cyclic polypeptides using a highly reproducible method.
As an example of the usefulness and importance of the MIC wild-type distributions, five isolates with an MIC of 8 mg/liter for capreomycin would be regarded as susceptible according to the present critical concentration (10 mg/liter) in Middlebrook 7H10 but were defined as non-wild type according to the ECOFF (4 mg/liter) (Fig. (Fig.1d).1d). The fact that these strains were resistant to amikacin and kanamycin makes it questionable whether they are fully accessible for treatment with standard doses of capreomycin. This is also supported by a calculated fCmax/MIC value close to zero at a MIC of 8 mg/liter for capreomycin (Table (Table4).4). Although we used Cmax levels in the lower range and there are variations in pharmacokinetics among individuals, it seems that the current critical concentration in 7H10 for capreomycin is too high. Truly resistant isolates may thus be reported as susceptible, which subsequently can lead to a susceptibility pattern with an overestimation of therapeutic options and in the end to the development of resistance to other drugs. We suggest that the clinical breakpoint (critical concentration) for capreomycin in Middlebrook 7H10 should be revised to 4 mg/liter.
Both MIC data as such and PD indices were more favorable for amikacin and viomycin than kanamycin and capreomycin. Although the pharmacodynamic target for M. tuberculosis is unknown, using the commonly accepted target for Gram-positive and Gram-negative bacteria of an fCmax/MIC value of ≥8 to 10 (10, 18), we could show that wild-type isolates below the ECOFF for amikacin, streptomycin, and viomycin were readily covered (Table (Table4).4). This supports the validity of our data, considering that streptomycin is one of the few drugs for which clinical outcome data in relation to MIC levels and critical concentrations are available (7). Of particular importance for the aminoglycosides, therapeutic drug monitoring is possible by MIC determination of the M. tuberculosis isolate combined with Cmax determinations and is available in most reference hospitals.
Today, critical concentrations for 7H10 are available for kanamycin, streptomycin, and capreomycin, but no critical concentrations are defined for amikacin and viomycin (25). According to our data, we suggest clinical breakpoints (critical concentrations) for amikacin and viomycin at the tentative ECOFFs (1 mg/liter for amikacin and 2 mg/liter for viomycin). It is possible, however, that adding MIC data for more strains using data from several laboratories could shift the suggested ECOFFs one MIC dilution step upwards. When we compared the MIC levels for amikacin obtained by Middlebrook 7H10 and Bactec 960 MGIT of nine M. tuberculosis strains, we found good agreement, indicating that the ECOFFs for both methods are likely to be similar. From this perspective, it is interesting to note that there was a large span of critical concentrations used in a survey of supranational reference laboratories for capreomycin in B460 (from 1.25 to 10 mg/liter) and that there was an 8-fold difference in the critical concentrations between the 7H10 (10 mg/liter) and Bactec 460 (1.25 mg/liter) methods, whereas there are no or very small differences for streptomycin (2 mg/liter versus 2 mg/liter) or kanamycin (4 mg/liter versus 5 mg/liter) (12, 25). If wild-type MIC distributions had been used to define these breakpoints, such variations could have probably been avoided.
A novel genotyping test based on known resistance mutations is reported to have an 85% sensitivity to detect resistance to amikacin and capreomycin in the rrs gene (5). However, these tests are based on known mutations found on well-characterized resistant strains, and, considering the increasing resistance against the injectable drugs, it is of importance to have a drug susceptibility method which could screen for strains with MICs that are higher than the ECOFF in order to detect and characterize novel resistance mutations.
The number of aminoglycoside- and cyclic polypeptide-resistant strains in our study is limited, since the primary aim was to define the wild-type MIC distributions and the ECOFFs. Still, regarding cross-resistance, it is interesting to note that two isolates that were clearly below the ECOFF and the critical concentration for streptomycin were resistant to amikacin, kanamycin, and capreomycin. These data underline that streptomycin should be tested with and considered a treatment alternative for XDR TB, even if the isolate is classified as resistant to other aminoglycosides and cyclic polypeptides. The seven isolates that were resistant to capreomycin, kanamycin, and amikacin according to the ECOFFs were all susceptible to viomycin, implying that cross-resistance is not as common as previously reported (16). Five of these isolates had a MIC higher than the ECOFF (4 mg/liter) for capreomycin but would have been classified as susceptible using the recommended critical concentration in Middlebrook 7H10 medium (10 mg/liter).
Compared to capreomycin, viomycin is not as well described in the literature as a treatment alternative for TB (6, 16, 20). Previous publications report that viomycin shares the major side effects of other aminoglycosides, such as nephro- and ototoxicity (20). Some authors claim that the nephrotoxicity might limit its use and recommend intermittent regimens of 1 or 2 g three times a week (20). The side effects should be possible to avoid by careful drug monitoring of the patient, which could be important in the treatment of drug-resistant TB (19, 20). Our in vitro data indicate that viomycin could be a good choice for treatment of aminoglycoside drug-resistant TB, as overall resistance rates were low and cross-resistance to other injectable drugs among the M. tuberculosis strains tested was very limited.
In conclusion, wild-type MIC distributions of aminoglycosides and cyclic polypeptides were determined, including tentative epidemiological wild-type cutoffs. Our data clearly suggest that the critical concentration for capreomycin should be reconsidered and that viomycin and streptomycin should be considered treatment options for MDR and XDR TB, even in the case of resistance to other class representatives.
This work was supported by FORSS (The Research Council of Southeast Sweden), the Karolinska University Hospital, and EC project LSHP-CT-2007-037912 (FAST-XDRDETECT).
Published ahead of print on 17 March 2010.