|Home | About | Journals | Submit | Contact Us | Français|
Antimicrobial susceptibility testing (AST) of clinical isolates of Nocardia is recommended to detect resistance to commonly used antimicrobial agents; such testing is complicated by difficulties in inoculum preparation and test interpretation. In this study, six laboratories performed repetitive broth microdilution testing on single strains of Nocardia brasiliensis, Nocardia cyriacigeorgica, Nocardia farcinica, Nocardia nova, and Nocardia wallacei. For each isolate, a total of 30 microdilution panels from three different lots were tested at most sites. The goal of the study was to determine the inter- and intralaboratory reproducibility of susceptibility testing of this group of isolates. Acceptable agreement (>90% agreement at ±1 dilution of the MIC mode) was found for amikacin, ciprofloxacin, clarithromycin, and moxifloxacin. After eliminating MIC values from single laboratories whose results showed the greatest deviation from those of the remaining laboratories, acceptable agreement was also found for amoxicillin-clavulanic acid, linezolid, minocycline, and tobramycin. Results showed unsatisfactory reproducibility of broth microdilution testing of ceftriaxone with N. cyriacigeorgica and N. wallacei, tigecycline with N. brasiliensis and N. cyriacigeorgica, and sulfonamides with N. farcinica and N. wallacei. N. nova ATCC BAA-2227 is proposed as a quality control organism for AST of Nocardia sp., and the use of a disk diffusion test for sulfisoxazole is proposed as a check of the adequacy of the inoculum and to confirm sulfonamide MIC results.
Nocardia species are ubiquitous in the environment and have been implicated in a variety of human infections, especially in immunocompromised patients. In these patients, infection most frequently begins in the lungs, but can quickly disseminate to nearly every organ of the body. Nocardia can also cause disease in the immunocompetent host, in whom it is usually introduced through traumatic injury (2).
Nocardia species are known to vary in their susceptibilities to various antimicrobial agents, and some species, especially Nocardia farcinica and members of the Nocardia transvalensis complex, have been shown to be particularly resistant to commonly used antimicrobial agents. Therefore, susceptibility testing should be performed on any Nocardia isolate considered to be of possible clinical significance.
In 2003, the Clinical and Laboratory Standards Institute (CLSI) published an approved standard (M24A) for susceptibility testing of mycobacteria, nocardiae, and other aerobic actinomycetes, which specifies the broth microdilution method as the recommended procedure for susceptibility testing of aerobic actinomycetes (6). This document has recently been updated (5). Several years of experience in testing Nocardia isolates at the University of Texas Health Science Center and at the National Institutes of Health have demonstrated difficulties that may occur in inoculum preparation due to the propensity for clumping demonstrated by these organisms. In addition, determination of susceptibility endpoints for some drugs has been problematic due to differences in growth characteristics among members of this genus.
The goal of this study was to evaluate the inter- and intralaboratory reproducibility of susceptibility testing of a defined group of Nocardia species and to evaluate best practices for inoculum preparation, endpoint determination, and result analysis. An additional goal of the study was to identify a Nocardia isolate that would act as a reference for growth characteristics of Nocardia in the microdilution panels.
Six laboratories performed replicate testing with all test organisms and all panel lots. Results from an additional site (The University of Texas Health Science Center at Tyler, TX) were considered the reference values for study results for N. brasiliensis, N. cyriacigeorgica, N. farcinica, and N. nova because of that laboratory's extensive clinical and laboratory experience with Nocardia isolates. (Data from this laboratory were not included in the data analysis, as the laboratory was unable to perform all of the replicate testing.)
Five clinical isolates of Nocardia species, previously tested for antimicrobial susceptibility by the broth microdilution method at the University of Texas Health Science Center at Tyler, TX, were selected for study based on various patterns of susceptibility or resistance. These included a single isolate each of N. brasiliensis, N. cyriacigeorgica, N. farcinica, N. nova, and N. wallacei (a representative of the N. transvalensis complex). These isolates were representatives of the most commonly isolated Nocardia species and were members of species that showed various susceptibility patterns (2). The identifications of all isolates were confirmed by 16S rRNA and secA1 gene sequencing (7, 8). Upon receipt in each laboratory, isolates were subcultured onto sheep blood agar and then frozen in tryptic soy broth with 15 to 20% glycerol at −70°C until thawed for use. Staphylococcus aureus ATCC 25923 and Pseudomonas aeruginosa ATCC 27853 were used as the quality control (QC) strains for the disk diffusion testing, and P. aeruginosa ATCC 27853 and S. aureus ATCC 29213 were used as the quality control strains for microdilution testing.
The antimicrobial agents evaluated by the microdilution and the dilution ranges tested are listed in Table 1. The panel included only one dilution series for each antimicrobial agent incorporated into the panel. The positive control wells contained no antimicrobial agent, and the negative control remained uninoculated. Three lots of custom-made, frozen microdilution panels, each containing a different lot of Mueller-Hinton broth (ThermoFisher Scientific, Cleveland, OH), were stored at −70°C until use. Drug concentrations were 2-fold serial dilutions in cation-adjusted Mueller-Hinton broth within the ranges listed in Table 1; the total volume in each well was 100 μl. In addition, ceftriaxone, imipenem, and sulfisoxazole were tested by disk diffusion testing on Mueller-Hinton agar plates (Becton Dickinson, Franklin Lakes, NJ). These three drugs were chosen for agar disk diffusion in order to address discrepancies and/or difficulties in interpretation of broth microdilution endpoints that have previously been noted at the University of Texas Health Science Center and the National Institutes of Health.
Each isolate was subcultured to a sheep blood agar plate (Becton Dickinson), and susceptibility testing was performed after 3 to 7 days of incubation at 35 ± 2°C in ambient air. Frozen isolates were subcultured twice before testing. A heavy suspension of organism was prepared in approximately 0.5 ml sterile water in a microcentrifuge tube. The inoculum was emulsified using a pellet pestle (Kimble Chase, Vineland, NJ) and the suspension was allowed to sit undisturbed (approximately 10 to 15 min) until large clumps had settled to the bottom of the tube, leaving a smooth supernatant. Several drops of the supernatant were added to tubes containing 5 ml of sterile water to prepare a suspension approximating a 0.5 McFarland turbidity standard. The suspensions were vortexed, and the turbidity of each suspension was checked using a nephelometer or spectrophotometer and adjusted as necessary to fall within the turbidity values for a 0.5 McFarland suspension as specified for each nephelometer or spectrophotometer used. The turbidities of the N. cyriacigeorgica, N. farcinica, and N. nova suspensions were adjusted to fall in the middle of the range appropriate for a 0.5 McFarland suspension, while the suspensions of N. brasiliensis and N. wallacei were adjusted to fall at the upper end of the range expected for a 0.5 McFarland suspension. (Initial studies performed at the NIH prior to the initiation of this study had indicated that clumps of N. brasiliensis and N. wallacei were difficult to break up, and the resulting suspension contained fewer CFU than the suspensions of N. cyriacigeorgica, N. farcinica, and N. nova.) After standardization, 1 ml of the adjusted suspension was added to a tube containing 29 ml of sterile water (bioMérieux, Durham, NC), and the tube was inverted 3 times to mix; the 0.5 McFarland suspension was set aside for disk diffusion testing.
The entire contents of the 30-ml suspension were added to the trough portion of an inoculation tray set. Microdilution panels were inoculated using a pronged inoculator (Evergreen Scientific, Los Angeles, CA), delivering 10 μl of the organism suspension to each well. The purity of the suspension was checked by subculture from the trough to a sheep blood agar plate, and a colony count was performed as described below. Microdilution panels were labeled, covered with plastic film, and incubated in plastic bags at 35 ± 2°C in ambient air. All panels were examined after 3 days of incubation; if growth was sufficient in the growth control well (growth score of approximately 2+ to 3+, as illustrated in Fig. 1), MIC values were recorded using the guidelines shown in Fig. 1. Incubation was continued up to 5 days for those isolates that showed insufficient growth in the positive control well at 3 days. All panels evaluated in this study showed sufficient growth after 3 to 5 days of incubation.
Colony counts were performed at the time of panel setup for one of the Nocardia replicates and for each QC strain. After susceptibility panel inoculation, 10 μl was removed from one of the two positive control wells in the panel and transferred to a tube containing 10 ml of sterile water. The suspension was vortexed, and 100 μl was removed and spread on a sheep blood agar plate. This procedure was done in duplicate. Plates inoculated with Nocardia sp. were incubated 3 to 5 days at 35 ± 2°C in ambient air until growth sufficient for counting colonies was achieved. The number of colonies seen on each of the two plates was recorded; CFU/ml was calculated as 104 times the average number of colonies on each plate. The target inoculum colony count was 1.0 × 105 to 5.0 × 105 CFU/ml in each well (10 to 50 colonies on each plate.)
Disk diffusion testing was performed from each standardized 0.5 McFarland suspension described above, using standard techniques (3). Antimicrobial disks (ceftriaxone, 30 μg; imipenem, 10 μg; and sulfisoxazole, 250 μg) (Becton Dickinson) were placed on 150-mm-diameter Mueller-Hinton plates (Becton Dickinson) using a template designed to provide maximum distance between disks. Mueller-Hinton plates were placed in plastic bags and incubated at 35 ± 2°C in ambient air for 3 to 5 days. Interpretive categories for disk diffusion results for Nocardia isolates were based on results obtained by Wallace and Steele using multiple strains of several species with the various antimicrobials (17). In these initial studies using Nocardia MIC values to determine zone diameter breakpoints, insufficient data were available to determine an “intermediate” susceptibility to sulfisoxazole. Zone diameters listed in Table 2 for the “intermediate” interpretive category for sulfisoxazole are therefore considered tentative.
QC testing was performed each week Nocardia isolates were tested, using the two S. aureus strains and one P. aeruginosa strain listed above. One suspension approximating a 0.5 McFarland turbidity standard was prepared for each strain, and turbidity was standardized using a nephelometer or spectrophotometer. Microdilution panels were inoculated as described above, except that only one panel from each lot was tested with both S. aureus ATCC 29213 and P. aeruginosa ATCC 27853. Disk diffusion testing was performed as described above using the same Pseudomonas strain and S. aureus ATCC 25923. Microdilution panels and disk diffusion plates were incubated in ambient air at 35 ± 2°C for 24 h. Quality control was considered acceptable if the results were within ranges recommended by the CLSI (4). The results for any antimicrobial agent that fell outside the recommended range were not utilized for the Nocardia isolates tested that week.
All susceptibility panels, antimicrobial disks, and media (other than some sheep blood agar plates) for the study were provided from the same sets of lots to minimize variation resulting from interlot differences as much as possible.
For each Nocardia isolate, susceptibility testing was performed on each of 3 separate testing days. On testing days 1 and 2, three standardized suspensions approximating a 0.5 McFarland suspension were prepared from the initial heavy suspension; one 30-ml suspension was prepared from each of the three standardized 0.5 McFarland suspensions (three 30-ml suspensions altogether), and each was transferred to an inoculation trough. One microdilution panel from each of the three lots was inoculated from each of the 30-ml suspensions (as per the procedure described above), for a total of 9 panels inoculated (3 from each lot). The pronged inoculator was changed after the inoculation of each panel. On testing day 3, four suspensions approximating a 0.5 McFarland suspension were prepared and, following the procedure outlined above, a total of 12 panels were inoculated (4 from each of the 3 lots). Again, the pronged inoculator was changed after inoculation of each panel. For each organism, a total of 30 panels from 3 different lots were tested at most sites. Some sites were not able to complete all replicates.
MIC values, disk diffusion zone diameters, and colony count results were recorded at each site on a standardized form, and overall results were compiled by one of the investigators (P.S.C.).
For most antimicrobial agents, the MIC was considered to be the lowest concentration of the drug that inhibited visible growth. Growth in each well was graded, as shown in Fig. 1. Except for moxifloxacin, MIC breakpoints were those described by the CLSI in 2003 (5). The breakpoints for moxifloxacin were those described by the CLSI in 2011 (4). There are no breakpoints for tigecycline. MIC breakpoints are listed in Table 1.
For the sulfonamides (sulfamethoxazole and trimethoprim-sulfamethoxazole), the endpoint was the well corresponding to 80% inhibition of growth. For these drugs, if the well with the lowest concentration of the drug showed better growth than the control, that well was used as the reference well for determining the MIC, rather than the growth control well.
Zones of inhibition were measured in millimeters, and zones of both partial (fine growth outside a smaller, complete zone of inhibition) and complete inhibition were recorded. Except for sulfisoxazole, for which the zone diameter recorded was the point of 80% inhibition of growth, the diameter of the complete zone of inhibition was used for disk interpretation. Zones for imipenem when tested with N. farcinica showed “microcolonies” within the zone of inhibition; ignoring these colonies, the zone size was considered to be the diameter at the point of heavy growth (B. Brown-Elliott and R. J. Wallace, Jr., personal communication). Disk diffusion tests were interpreted according to methods outlined in CLSI document M100 (4), using breakpoints established by Wallace and Steele (17) (Table 2).
Quality control results obtained by 4 testing sites were examined. For each antimicrobial-organism combination, the mode MIC value for each lot was determined. The mode values for each antimicrobial-organism combination for each of the 3 lots were compared, and mode values that differed by more than ±1 dilution were determined to indicate lot-to-lot variation.
For each of the 6 testing sites, MIC results for each antimicrobial-organism combination obtained in a specific run (defined as all tests for a specific antimicrobial-organism combination set up on a specific day by an individual laboratory) were compared, and the mode MIC value for that antimicrobial-organism combination in that run was determined. Mode values for the three runs were compared, and runs with modes that varied more than ±1 dilution from the other runs were considered to have considerable variation.
Data were imported into SAS (Statistical Analysis Systems) version 9.2 and were analyzed separately for each Nocardia species and each antimicrobial combination. For each species and antimicrobial combination, the following values were determined: (i) the MIC dilution mode and range, (ii) the percentage of MIC dilutions that encompassed the mode ± 1 2-fold dilution of the mode, (iii) the percentage of MIC dilutions that encompassed the mode ± 2 2-fold dilutions of the mode, and (iv) the percentage of agreement based on interpretive category.
For antimicrobial-organism combinations that showed less than 90% agreement at mode ± 1 dilution among testing sites, results from each site were systematically eliminated to explore the effect of the results of each testing site on agreement among the remaining sites. The percentages of MIC dilutions that encompassed the mode ± 1 2-fold dilution of the mode among the remaining five sites were recalculated with each elimination.
To examine intralaboratory reproducibility, two measures of lack of reproducibility were analyzed: (i) the percentage of all antimicrobial-organism combination results that were outside the mode ± 1 dilution and (ii) the percentage of antimicrobial-organism combinations with readings that spanned more than 3 dilutions. To compare the percentages of all antimicrobial-organism combination results that were outside the mode ± 1 dilution among the 6 laboratories, a 2-by-6 Fisher's exact test was performed. When the 2-sided P value was significant, 2-by-2 Fisher's exact tests were performed for all 15 pairwise comparisons. The resulting P values were then adjusted by the Bonferroni method. To compare the percentages of all antimicrobial-organism combination readings spanning >3 dilutions among the 6 laboratories, a Cochran's Q test was performed. When the P value was significant, a McNemar's test for matched pairs for all 15 pairwise comparisons was performed; the resulting P values were adjusted by the Bonferroni method. P values less than 0.05 were considered statistically significant.
Summaries of MIC results for all antimicrobial-organism combinations are shown in Tables 3 to to7.7. Results for the following antimicrobials showed >90% agreement within the mode ± 1 dilution for all laboratories for each Nocardia species tested: amikacin, ciprofloxacin, clarithromycin, and moxifloxacin. For linezolid (with N. brasiliensis), amoxicillin-clavulanic acid (with N. nova), and tobramycin (with N. nova), removal of results from laboratory 6 resulted in agreement at the mode ± 1 dilution of 90.0, 94.9, and 96.4%, respectively (results not shown). For minocycline, removal of results from laboratory 5 for N. brasiliensis resulted in 91.3% agreement at the mode ± 1 dilution.
Good agreement at the mode ± 1 dilution was seen when testing N. brasiliensis, N. farcinica, and N. nova. With N. cyriacigeorgica, cumulative ceftriaxone results spanned all dilutions tested. N. cyriacigeorgica is expected to be susceptible to ceftriaxone (2); however, the mode value obtained in this study was in the resistant range (64 μg/ml) (Table 4), representing only 28.1% of all results (data not shown). Evaluation of the effect on result agreement by eliminating ceftriaxone MIC results from one site at a time did not improve the percentage of agreement to >90% at either the mode ± 1 dilution (best agreement, 70.3%) or the mode ± 2 dilutions (best agreement, 80.7%) (data not shown). Results based on interpretive category show a nearly equal distribution of susceptible, intermediate, and resistant results (34.3, 31.4, and 34.3%, respectively) (Table 4). Disk diffusion testing of N. cyriacigeorgica with ceftriaxone (56 total trials) also gave variable results, with 21.4% of results interpreted as susceptible and 71.4% interpreted as resistant (Table 2). The reference laboratory result for N. cyriacigeorgica with ceftriaxone was susceptible, with MICs ranging from 8 to 16 μg/ml (data not shown).
For ceftriaxone with N. wallacei, cumulative results from all laboratories again spanned all dilutions tested. N.wallacei is expected to be ceftriaxone susceptible, and the mode value was 8 μg/ml (60.6% agreement at the mode ± 1 dilution), in the susceptible range (Table 7). Evaluation of the effect on result agreement by the elimination of ceftriaxone results from one laboratory at a time did not improve the percentage of agreement to >90% at the mode ± 1 dilution (best agreement, 72.6%), but the elimination of results from laboratory 2 improved agreement at the mode ± 2 dilutions to 94.0%. By interpretive category, 47.8 and 37.2% of results were in the susceptible and intermediate categories, respectively. For disk diffusion testing, laboratory 1 reported all tests as resistant; all other laboratories reported susceptible or intermediate results. There are no reference laboratory results for N. wallacei with ceftriaxone.
Good agreement at the mode ± 1 dilution was seen when testing N. brasiliensis and N. nova. While the MICs for imipenem testing of N. cyriacigeorgica showed only 75.8% agreement at the mode ± 1 dilution, results spanned only 3 dilutions and there was 100% agreement of susceptibility among laboratories by interpretive category (Table 4). Elimination of MIC results from either laboratory 1 or 2 resulted in 100% agreement at the mode ± 1 dilution. Agreement at the mode ± 2 dilutions was 100%. N. cyriacigeorgica is expected to be susceptible to this drug, and reference laboratory MIC results ranged from 1 to 2 μg/ml (susceptible). Disk diffusion testing gave reproducible results, with 98.2% of 56 tests reported as susceptible.
For imipenem testing with N. farcinica, 97.8% of results were in the susceptible range (4 dilutions) and only 2.2% of results were in the intermediate range. However, agreement at the mode ± 1 dilution was 73.9% and was not improved above 87.3%, even with elimination of results from one laboratory at a time. Agreement at the mode ± 2 dilutions was 100%. Reference laboratory results for this isolate ranged from 1 to 2 μg/ml (susceptible). Results for disk diffusion with imipenem showed 85.5% agreement of a susceptible interpretation. N. farcinica has been shown to exhibit variable susceptibility to imipenem (2).
For N. wallacei, all imipenem results clustered among 3 dilutions in the intermediate and resistant range, with the mode value being greater than the highest dilution tested (>16 μg/ml) (Table 7). Elimination of results from either laboratory 2 or laboratory 6 resulted in 100% agreement at the mode ± 1 dilution. At the mode ± 2 dilutions, all results were at 100% agreement. In contrast, disk diffusion testing showed 56.4% of tests to be susceptible and 43.6% resistant (Table 2). N. wallacei has been shown to exhibit varying susceptibility to imipenem (2). There was no reference laboratory result for this drug with this N. wallacei strain.
Good agreement at the mode ± 1 dilution was seen when testing N. farcinica, N. nova, and N. wallacei. For N. brasiliensis, results spanned 8 dilutions and four laboratories showed results that were outside the mode ± 1 dilution. Analysis of the data by elimination of results from one laboratory at a time did not improve the percentage of agreement above 80.7%; agreement at the mode ± 2 dilutions was 90% or greater. There are no established interpretive categories for this drug, and the expected susceptibility of N. brasiliensis to tigecycline has not been determined. Reference laboratory results showed that the MIC for this isolate ranged from 0.12 to 0.5 μg/ml.
For N. cyriacigeorgica, results spanned a wide range of dilutions. Agreement at the mode ± 1 dilution was 42.1% (Table 4) and did not improve appreciably even with elimination of results from one laboratory at a time. Agreement at the mode ± 2 dilutions was 65.7%. Agreement by interpretive category for this drug is unknown due to a lack of breakpoints, and the expected susceptibility testing result for this species has not been established. Reference laboratory results for N. cyriacigeorgica with tigecycline ranged from 0.5 to 1.0 μg/ml, 20 to 3 dilutions less than the mode MIC obtained by laboratories participating in the study.
Good agreement at the mode ± 1 dilution was seen when testing N. brasiliensis, N. cyriacigeorgica, and N. nova. A total of 98.4% of disk diffusion results for N. brasiliensis with sulfisoxazole were susceptible. For N. cyriacigeorgica and N. nova, 57.1% and 58.6%, respectively, of disk diffusion results were in the susceptible range with the remainder in the “intermediate” range (Table 2) (see Materials and Methods for discussion of the intermediate range for sulfonamides).
For N. farcinica, the mode MIC was 64 μg/ml (Table 5) in the resistant range. Agreement at the mode ± 1 dilution was 72.8% and did not improve above 82.0% agreement with elimination of results from one laboratory at a time. Agreement at the mode ± 2 dilutions was 95.0%. N. farcinica is expected to be susceptible to sulfamethoxazole, but based on interpretive category, results were equally divided between susceptible (50%) and resistant (50%). By disk diffusion testing, 93.6% of results were in the susceptible range and 6.4% in the “intermediate” range (Table 2). Reference laboratory results showed this N. farcinica isolate to be susceptible to sulfamethoxazole, with MICs ranging from 16 to 32 μg/ml.
For N. wallacei, results spanned all dilutions tested, with agreement of 52.9 and 75.0% at the mode ± 1 and 2 dilutions, respectively (Table 7). Elimination of results from one laboratory at a time increased the percentage of agreement at the mode ± 1 dilution to no more than 71.8%. Interestingly, only elimination of results from laboratory 3 resulted in >90% agreement at the mode ± 2 dilutions. N. wallacei is expected to be susceptible to sulfamethoxazole, but only laboratory 3 reported all MICs in the susceptible range. Overall, 53.5% of results were interpreted as susceptible and 46.5% interpreted as resistant. Disk diffusion testing showed 81.8 and 9.1% of results in the susceptible and intermediate ranges, respectively (Table 2); 9.1% were in the resistant range. There are no reference laboratory results for this N. wallacei isolate with sulfamethoxazole.
As with sulfamethoxazole, good agreement at the mode ± 1 dilution was seen when testing N. brasiliensis, N. cyriacigeorgica, and N. nova. With trimethoprim-sulfamethoxazole for N. farcinica, the mode MIC was 0.5 (trimethoprim)/9.5 (sulfamethoxazole) μg/ml (Table 5), well within the susceptible range. However, MICs from the 6 testing sites spanned all dilutions of that range. Agreement at the mode ± 1 dilution was 68.6% and did not improve above 75.3% agreement with elimination of results from one laboratory at a time. Agreement at the mode ± 2 dilutions was 92.9%. N. farcinica is expected to be susceptible to trimethoprim-sulfamethoxazole, and by interpretive category, 92.9% of isolates were within the susceptible range, with only 7.1% in the resistant range. The reference laboratory results for N. farcinica with trimethoprim-sulfamethoxazole ranged from 0.5/9.5 to 2/38 μg/ml, all within the susceptible range.
For N. wallacei, the mode MIC was 2/38 μg/ml (Table 7) and laboratories reported results ranging from ≤0.25/4.8 to 16/304 μg/ml. Agreement at the mode ± 1 dilution was 80.2% and did not improve to above 90% with elimination of results from one laboratory at a time. Agreement at the mode ± 2 dilutions was 97.1%. By interpretive category, 72.1% of results were reported as susceptible and 27.9% as resistant. N. wallacei is expected to be susceptible to trimethoprim-sulfamethoxazole. There were no reference laboratory results for trimethoprim-sulfamethoxazole with N. wallacei.
Achievement of the target inoculum concentration (1 × 105 to 5 × 105 CFU/ml) was species dependent, with suspensions of N. cyriacigeorgica, N. farcinica, and N. nova most frequently containing the desired number of CFU (71.4, 76.2 and 74.1%, respectively) (data not shown). Inocula of N. brasiliensis and N. wallacei deviated most from the target organism concentration, with only 22.6 and 57.1% of inocula, respectively, achieving that concentration.
The results of QC testing of all drugs from the three lots were examined from four of the six participating laboratories. Analysis showed that in all but 3 of 54 drug/lot tests, the mode values for replicates at each site with a particular drug tested by each of the three lots were within ±1 dilution of each other (data not shown). The three instances of nonagreement included three different drugs at 2 sites (amikacin from laboratory 1 and ceftriaxone and moxifloxacin from laboratory 4). Because no one drug from any lot varied consistently among the sites, the various lots were considered to be functionally identical.
The examination of run-to-run variability within each laboratory showed minimal variation over the course of the study for 5 of the 6 testing sites (data not shown). Sites 1 through 5 reported zero (1 site), one (3 sites), or two (1 site) instances of among-run variability. This variability was seen with several different antimicrobial agents and species, with no consistent drug or organism showing more variation than another. Laboratory 6 showed 8 instances of run-to-run variation with a variety of antimicrobial agents and Nocardia species.
Results from individual laboratories were evaluated for two measures of laboratory reproducibility (Table 8). Analysis of the percentage of readings outside the mode ± 1 dilution reported by each laboratory ranged from 3.0% of readings to 14.4%. Results from laboratory 3 were significantly more reproducible than those from all other sites (P < 0.001), while results from laboratories 5 and 6 were significantly less reproducible than those from all other sites, but not significantly different from each other (P < 0.001) (Table 8). The percentage of readings spanning more than 3 dilutions was also determined, and values ranged from 3.1% to 36.9%. The range of results obtained from laboratory 6 was significantly greater than the ranges reported by laboratories 1 through 4 (P < 0.01) and also greater than that reported by laboratory 5 (P = 0.035).
Antimicrobial susceptibility testing (AST) of clinically significant isolates of Nocardia species is recommended because certain isolates of this genus differ in susceptibility to commonly used antimicrobial agents (2). Generally, susceptibility or resistance to certain antimicrobial agents can be predicted for the most frequently isolated species of clinical significance, including N. brasiliensis, N. farcinica, N. cyriacigeorgica, N. nova, and the N. transvalensis complex (represented in this study by N. wallacei) (9, 16, 18, 19). However, for other newly described species and some strains of the more common species, susceptibility patterns are unknown or are not predictable (2). The broth microdilution method is the recommended method for AST of Nocardia species (5).
One of the stated goals of this study was to assess the intra- and interlaboratory reproducibility of broth microdilution susceptibility testing of Nocardia isolates. Results showed that the level of reproducibility was partly related to the antimicrobial agent being tested.
There was acceptable reproducibility (≥90% agreement at the mode ± 1 dilution) among all laboratories for several antimicrobial agents, including amikacin, ciprofloxacin, clarithromycin, and moxifloxacin. These drugs gave clear and unambiguous endpoints that were easily interpreted by all laboratories. Selective removal of results from laboratories 5 and 6 resulted in acceptable reproducibility for amoxicillin-clavulanic acid, linezolid, minocycline, and tobramycin.
There were differences among testing sites in the extents of variation of their own results from the modal result. One laboratory (laboratory 3) reported significantly fewer readings outside the mode ± 1 dilution for all antimicrobial-organism combinations compared to the other 5 testing sites, and 2 laboratories (laboratories 5 and 6) reported significantly more readings outside the mode ± 1 dilution compared to the other testing sites (Table 8). In addition, one testing site (laboratory 6) reported a significantly higher number of readings spanning more than 3 dilutions (for the 65 antimicrobial-organism combinations) than the remaining laboratories (Table 8). The reasons for the lack of reproducibility reported by these laboratories are unclear. Endpoint determination of some antimicrobial agents was more difficult, possibly related to attributes of the agent itself, to the particular isolate tested, or to complexities of the testing methodology.
Our assessment of lot-to-lot variation in the microdilution plates indicated that panel lots were functionally identical and did not contribute to result variation. Likewise, run-to-run variation was minimal, except for laboratory 6, which reported 8 instances of variations among runs.
Considerable deviation from the mode value was seen for several antimicrobial agents, including ceftriaxone, imipenem, sulfamethoxazole, trimethoprim-sulfamethoxazole, and tigecycline. MICs for ceftriaxone were most problematic for N. cyriacigeorgica and N. wallacei, with inadequate agreement (<90%) even at the mode ± 2 dilutions. In addition, disk diffusion results were also ambiguous for these two species that are considered susceptible to this drug by in vitro MIC testing (2). It is unclear if these discrepancies were due to inadequate inocula or to interpretation inconsistencies due to the growth characteristics of these isolates. Given the wide variability of results from the various laboratories, it appears that the broth microdilution method as described may not allow reliable results for ceftriaxone to be obtained with N. cyriacigeorgica and N. wallacei. The addition of disk diffusion testing was not useful for resolving testing inconsistencies.
MIC results for imipenem with N. cyriacigeorgica and N. farcinica were only in agreement at the mode ± 2 dilutions; both species are considered susceptible to this drug (2). However, there was complete agreement among test sites by interpretative categories, and disk diffusion tests for N. cyriacigeorgica were reproducibly susceptible. For N. farcinica, the disk diffusion discrepancies from the expected susceptible result may in part be due to the more rapid growth of this species compared to other species of Nocardia. Moreover, because of the characteristic smooth colony morphology of N. farcinica (smooth colonies suspend more efficiently and produce a more homogenous inoculum in broth than rough colonies), this species can easily be overinoculated when preparing the test inocula. In addition, imipenem disk diffusion results for N. farcinica may have been misinterpreted as resistant by one testing site due to the presence of tiny colonies within the zone of inhibition; these colonies are characteristic of this antimicrobial-organism combination and are not indicative of resistance (R. J. Wallace, Jr., and B. Brown-Elliott, unpublished observations). All other testing sites reported imipenem zone diameters within the susceptible or intermediate ranges when testing N. farcinica. Both MIC and disk diffusion results for N. wallacei with imipenem were not reproducible; this species shows variable susceptibility to this drug, possibly based on the difficulty in test interpretation or the difficulty in achieving consistent inoculum concentrations.
Sulfonamides, often in combination with other antimicrobial agents, remain the drugs of choice for treatment of nocardial infections because of the historical in vitro susceptibility of Nocardia isolates to these drugs and because of observed clinical effectiveness (12). In this study, there was considerable lack of reproducibility of sulfonamide MIC results, especially when testing N. farcinica and N. wallacei. Because of the unique growth characteristics of Nocardia species, the correct endpoint may be difficult to determine; this is most probably the reason for the discrepant sulfonamide MIC results observed in this study. Guidelines state that the endpoint is the well exhibiting 80% inhibition of growth compared to the positive control well (5). Wells containing low concentrations of sulfonamide may also be used as a comparison for 80% inhibition, as better growth at the lowest drug concentration compared to the single growth control is frequently observed (R. J. Wallace, Jr., and B. Brown-Elliott, unpublished observations). Determination of the correct endpoint is critical. Because of a lack of intermediate breakpoints for both sulfamethoxazole and trimethoprim-sulfamethoxazole, a 1-dilution endpoint difference can mean the difference between a susceptible and a resistant result. It should be noted that current CLSI guidelines recommend testing only trimethoprim-sulfamethoxazole; interestingly our results show more consistency with trimethoprim-sulfamethoxazole testing than with sulfamethoxazole alone.
Analysis of sulfonamide MIC results by interpretive category indicates that a considerable number of isolates were unexpectedly reported as resistant to these drugs. Of a total of 1,681 MIC results for both sulfamethoxazole and trimethoprim-sulfamethoxazole with all species, 1,436 (85.4%) isolates were reported as susceptible and 245 (14.6%) were reported as resistant (data not shown). In contrast, by disk diffusion, only 1.7% of 296 total Nocardia sulfisoxazole results were interpreted as resistant, with 78.7 and 19.6% reported as susceptible and “intermediate,” respectively. Because this resistance was seen inconsistently among testing sites and species, we conclude that “resistant” results are more an indication of inaccurate sulfonamide endpoint determination than of intrinsic resistance to the drugs tested.
Results of this study indicate that the use of the disk diffusion test for sulfonamides is useful for validation of MIC results (see below). In addition, because of the difficulty of interpretation of sulfonamide MICs, it is possible that susceptibility to sulfonamides may be more reliably predicted by disk diffusion testing. Zones of inhibition seen with disk diffusion may be easier to interpret; CLSI guidelines for interpretation of sulfonamides indicate that slight growth surrounding the disk (20% or less of the lawn of growth) should be disregarded and the more obvious margin of growth should be measured to determine the zone diameter (4).
In a recent retrospective study of 765 Nocardia isolates, Uhde et al. reported unusually high levels of in vitro sulfonamide resistance, as determined by MIC testing, with 61% of isolates showing resistance to sulfamethoxazole and 42% showing resistance to trimethoprim-sulfamethoxazole (14). As the authors note, the set of isolates may have included many that were unresponsive to the usual treatment provided. However, no data were available to those authors regarding the antimicrobial treatment or the clinical outcomes. Particularly, it is not known how many patients were treated successfully with sulfonamides. In contrast, in an additional retrospective study of sulfonamide susceptibility results of 552 recent clinical Nocardia isolates from various regions of the United States, Brown-Elliott et al. reported only 2% of isolates to have MICs reported indicating resistance to trimethoprim-sulfamethoxazole or sulfamethoxazole (1). Given the data presented by Brown-Elliott et al. and our own data demonstrating the difficulty of sulfonamide susceptibility test interpretation with these organisms, we are concerned that the in vitro results reported by these Uhde et al. may not accurately reflect in vivo responsiveness.
The reasons for poor agreement among laboratories at the mode ± 1 dilution for tigecycline with N. cyriacigeorgica and N. brasiliensis are unclear. At least for N. brasiliensis, inconsistencies may be due to the growth characteristics of this species and the difficulties associated with inoculum preparation.
A secondary goal of this study was to select a candidate clinical isolate to serve as a quality control organism for susceptibility testing of the Nocardia species. Although none of the isolates evaluated was a perfect candidate, N. nova was closest to optimal, with the most reproducible, although highly susceptible, results. Laboratories should obtain results similar to those obtained in this study with this organism (Table 6). The inclusion of this strain in the quality control battery is useful to show the particular attributes of Nocardia strains grown in the presence of antimicrobial agents. This strain has been deposited in the American Type Culture Collection (ATCC) as ATCC BAA-2227. Additional quality control organisms should also be tested, as outlined in the most recent CLSI document (5).
As with other organisms, inoculum preparation is an important factor in AST of Nocardia. Because of the clumping properties of these organisms, preparation of an adequately homogeneous organism suspension is especially challenging and important. In this study, we tested the use of pellet pestles to grind clumps of organisms in a small volume of water, thereby creating a dense suspension with fewer clumps. For N. cyriacigeorgica, N. farcinica, and N. nova, the supernatant resulting from this technique was adequately dense to allow preparation of a suspension corresponding to a 0.5 McFarland standard, as determined by nephelometer reading. For N. brasiliensis and N. wallacei, the clumps were more difficult to disperse, and the supernatant was less dense than with other species, and therefore a larger volume of supernatant was necessary to achieve the target suspension density. Interestingly, results obtained in this study for N. wallacei were the least reproducible with several antimicrobial agents, possibly because of the difficulty of inoculum preparation.
The CLSI procedure states a target organism concentration of 1 × 105 to 5 × 105 CFU/ml as the goal for accurate susceptibility testing (5). Our results show that the achievement of an adequate colony count was directly related to the ease with which the organism suspension was obtained. Overall, colony counts obtained with N. cyriacigeorgica, N. farcinica, and N. nova were more frequently in the expected range, while colony counts obtained with N. brasiliensis and N. wallacei were lower than expected (data not shown). Because in most cases, isolate identification is unknown at the time AST is initiated and because of the differences observed in colony count for the various species, inocula may best be prepared at the middle of the acceptable range of nephelometer readings for all organisms in an attempt to achieve an adequate organism concentration for all species.
The relationship between Nocardia concentration (as determined by colony count) and MIC results has not previously been explored. In this study, participating laboratories reported colony counts either within the target range or lower than the target range for each isolate tested and MIC interpretation discrepancies occurred regardless of the colony count reported (data not shown). Some laboratories that reported organism concentrations within the target range reported MICs 2 or more dilutions greater or less than the mode value. Likewise, some laboratories that reported colony counts less than the target range reported MICs that were within ±1 dilution of the mode value. This suggests that colony counts did not give a true representation of the inocula, and we recommend that they not be performed. Variables other than colony count appear to more significantly affect the breakpoint determination.
More experienced investigators recommend the use of the agar disk diffusion method in combination with the broth MIC to help determine the adequacy of the inoculum (17). Until more work is performed to optimize inoculum standardization, we recommend that an agar disk diffusion plate, preferably with a sulfisoxazole disk, be incorporated in the test setup. This can be prepared from the 0.5 McFarland suspension initially prepared for microdilution testing. If, after incubation, the growth on the Mueller-Hinton plate is confluent (inoculum too heavy), or if isolated colonies are present (inoculum too light), the entire test should be repeated with careful attention to inoculum density. An appropriate inoculum would show streaks of growth with spaces between the streaks (Fig. 2). Based on previous studies comparing zone diameter on disk diffusion testing to MIC results, this growth appearance is different from that expected from Kirby Bauer testing of other bacteria but has been determined to be indicative of adequate Nocardia inoculum (17). This may be due to the colony size and growth characteristics of Nocardia species compared to other bacteria.
As mentioned above, the use of disk diffusion with a sulfisoxazole disk is also useful to check the accuracy of the sulfonamide MIC, especially if the MIC reading indicates that the isolate is sulfonamide resistant. Results obtained in this study showed disk diffusion results to be more indicative of susceptibility for some organisms. A discrepancy between the sulfonamide MIC and disk diffusion interpretive category would indicate that the test should be repeated or sent to a referral laboratory for result confirmation. The disk diffusion result should not be reported; it is solely intended to verify the accuracy of the sulfonamide MIC result.
Comparison of MIC results with expected results for a particular species is especially useful for drugs that are difficult to test, which may give ambiguous results and which are potentially useful in a clinical situation (such as the sulfonamides and, if tested at all, ceftriaxone, for N. cyriacigeorgica and N. wallacei). MICs that differ from the expected results should be repeated to verify those results.
Currently no proficiency test service (such as that of the College of American Pathologists) is available specifically for Nocardia susceptibility testing. If a laboratory elects to perform in-house testing, test performance should be validated prior to offering the test, which is best accomplished through comparison of results with an experienced accredited reference laboratory. In addition, proficiency testing should be done at least twice per year either by comparing results with a reference laboratory or by repeating the evaluation of isolates previously tested.
While this study examined the reproducibility of susceptibility testing of Nocardia species, there are other factors unrelated to the susceptibility process itself that may make application to patient care more problematic. Knowledge of the mechanisms of drug resistance, particularly to generally effective agents such as the sulfonamides, can be very useful in clinical practice. Such knowledge would be particularly helpful in the case of drugs for which clinical effectiveness may be difficult to determine in vitro. (Again, the sulfonamides are an example.) Knowledge of specific resistance-conferring mechanisms such as inducible β-lactamase resistance in N. nova complex might explain some variable test results with β-lactams and eventually make rapid and unequivocal detection of resistance by molecular methods relatively simple (11, 13, 15). The detection of inducible β-lactamase was not addressed in this study. Given that such inducibility exists in the N. nova complex, caution should be exercised in the treatment of patients with isolates of N. nova complex reported to have MICs to amoxicillin-clavulanic acid in the susceptible range. Unfortunately, to date there has been a lack of data regarding resistance mechanisms in most species of Nocardia, making AST a continuing necessity.
Recent studies have shown that N. farcinica contains multiple genes relating to β-lactam, aminoglycoside, and macrolide resistance (10). These findings have led investigators to speculate on the presence of antimicrobial resistance mechanisms such as aminoglycoside-modifying enzymes seen in N. farcinica. These mechanisms may play a significant role in the finding of a wide range of results of MICs when testing individual strains of Nocardia. These mechanisms may, in part, also explain the apparent in vitro susceptibility of amikacin sometimes seen even with extended incubation in species known to be aminoglycoside resistant in vivo (e.g., N. transvalensis complex). (R. J. Wallace, Jr., and B. Brown-Elliott, unpublished observations).
Results obtained in this study illustrate the numerous difficulties that may be encountered in the setup and interpretation of broth microdilution testing of Nocardia species and highlight the variation in test interpretation that may occur even within a particular laboratory. These difficulties may influence the decisions of some laboratories to institute this test. However, one laboratory participating in this study (laboratory 2) had limited previous experience with broth microdilution testing of Nocardia isolates but reported only 6.8% of overall results outside the mode ± 1 dilution and only 6.2% of results spanning greater than 3 dilutions (Table 8). All results were read and interpreted by a single individual with no previous Nocardia susceptibility testing experience, who was carefully trained and who referred to the interpretive guidelines provided for the study. The experience of laboratory 2 suggests that a relatively inexperienced person who is well trained, meticulous, and conscientious and who is adequately supervised can perform and accurately interpret MIC results for Nocardia species.
In summary, strict attention to appropriate inoculum preparation, thorough training of technical staff, attention to detail by designated test readers, and willingness to seek help with troublesome interpretations may allow for more accurate Nocardia MIC result interpretation. Laboratories performing this test should become familiar with endpoint determinations, as illustrated in this document (Fig. 1) and in the current CLSI guidelines (5). Comparison of results with expected results for a particular species is especially useful for drugs that are difficult to test and which are potentially useful in a clinical situation (such as ceftriaxone, imipenem, tigecycline, and the sulfonamides). Any isolate that does not have the expected antimicrobial susceptibility pattern of that species should be retested or sent to a reference laboratory for result confirmation. The data from this study suggest that in vitro testing of N. cyriacigeorgica and N. wallacei (and possibly other Nocardia species) with ceftriaxone cannot currently be reliably performed using broth microdilution testing. In addition, sulfonamide testing by disk diffusion may provide more reliable results than testing by broth microdilution. The testing of appropriate quality control strains to confirm drug potency, the inclusion of the N. nova ATCC BAA-2227 strain as a reference for nocardial growth patterns in the microdilution panel, and regular proficiency testing will help ensure that accurate results are reported. In addition, we suggest that colony count plates not be inoculated and that a disk diffusion test for sulfisoxazole be incorporated into the test setup to check inoculum density and to confirm sulfonamide MIC results.
This work was supported in part by grants from Pfizer (New York, NY), Wyeth (Madison, NJ), ThermoFisher Scientific (Cleveland, OH), and BD (Franklin Lakes, NJ).
We thank Laura Doyle of the Cleveland Clinic for technical assistance, Tracy Dooley of the CLSI for organizational assistance, and Patrick R. Murray and Adrian M. Zelazny, NIH, for critically reviewing the manuscript.
The views expressed here are those of the authors and should not be construed as those of the U.S. Department of Health and Human Services.
Published ahead of print 4 January 2012