Discordance across several methods for drug susceptibility testing of drug-resistant Mycobacterium tuberculosis isolates in a single laboratory

Abstract

Given the increases in drug-resistant tuberculosis, laboratory capacities for drug susceptibility testing are being scaled up worldwide. A laboratory must decide among several endorsed methodologies. We evaluated 87 Mycobacterium tuberculosis isolates for concordance of susceptibility results across six methods: the L-J proportion method, MGIT 960 SIRE AST, Gene/Xpert MTB/RIF, GenoType MTBDRplus line probe assay, MycoTB MIC plate, and a laboratory-developed mycobacteriophage quantitative PCR (qPCR)-based method. Most (80%) isolates were multidrug resistant. Of the culture-based methods, the mycobacteriophage qPCR method was fastest, the L-J proportion method was the slowest, and the MGIT method required the most repeat testing (P < 0.05). For isoniazid (INH), 82% of isolates were susceptible by all methods or resistant by all methods, whereas for rifampin (RIF), ethambutol (EMB), and streptomycin (STR), such complete concordance was observed in 77%, 50%, and 51% of isolates, respectively (P < 0.05 for INH or RIF versus EMB or STR). The discrepancies of EMB and STR stemmed largely from diminished concordance of the MGIT EMB results (kappa coefficient range, 0.26 to 0.30) and the L-J STR result (kappa range, 0.35 to 0.45) versus other methods. Phage qPCR and the MycoTB MIC plate were the only methods that yielded second-line susceptibilities and revealed significant quantitative correlations for all drugs except cycloserine, as well as moderate to excellent kappa coefficients for all drugs except for para-aminosalicylic acid. In summary, the performance of M. tuberculosis susceptibility testing differs by platform and by drug. Laboratories should carefully consider these factors before choosing one methodology, particularly in settings where EMB and STR results are clinically important.

FIG 1

Turnaround times and repeat testing results of DST methods. A total of 87 isolates were tested across several DST methodologies for the first-line drugs INH, RIF, EMB, and STR. (Top) The time from culture to DST result was measured for each methodology (L-J, MGIT 960, Trek MycoTB, and phage qPCR). *, P < 0.05 for each comparison. (Bottom) Percentages of specimens or isolates that failed due to contamination or insufficient growth and that required repeat testing, enumerated by platform. *, P < 0.05 (for MGIT, significantly higher than Trek MycoTB, LPA, or GeneXpert); **, P < 0.05 (for GeneXpert, significantly lower than MGIT, L-J, or phage qPCR).

FIG 2

Discordance of methods of DST for drug-resistant M. tuberculosis isolates. Isolates of M. tuberculosis were tested for drug susceptibility using the L-J proportion method, MGIT 960, GeneXpert, Hain MTBDRplus, MIC plate (Trek Sensititre MycoTB), and a laboratory-developed rapid mycobacteriophage qPCR-based method. In total, we used 5 methods for evaluation of INH resistance or susceptibility, 6 for RIF, 4 for EMB, and 4 for STR. Isolates were categorized as resistant by all methods, resistant by the majority of methods, susceptible by all methods, susceptible by the majority of methods, or indeterminate (if results with the different methods were split (e.g., 3 resistant/3 susceptible for RIF, or 2 resistant/2 susceptible for EMB or STR). The gold standard result for an isolate was defined as the result that the majority of methods yielded (arrows pointing up or down). Blue, red, and green portions indicate isolates for which discrepant results occurred. *, P < 0.05 for rate of complete concordance (percent resistant by all methods + percent susceptible by all methods divided by total number of isolates) for INH and RIF versus EMB and STR.

FIG 3

Drug susceptibility testing discrepancies by drug and methodology. Discrepancies versus the consensus gold standard were enumerated for INH, RIF, EMB, and STR by methodology. Discrepancy rates for EMB were higher for MGIT 960 than with other methods (29%, versus 4 to 5% for other methods; P < 0.05). Discrepancies were further delineated as susceptible (boxed outline indicating the majority were susceptible) or resistant (dashed outline indicating the majority were resistant). There were 0 discrepancies for INH by L-J. nd, not done (GeneXpert only evaluates RIF and the Hain MTBDRplus method only evaluates INH and RIF).

FIG 4

Correlations between mycobacteriophage qPCR C_q and MIC results. Eighty-four M. tuberculosis isolates were tested with both the Sensititre MycoTB and mycobacteriophage qPCR methods for second-line drugs. The correlation of MIC values with C_q value is shown for the indicated drugs. Dashed lines indicate the concentration breakpoints used for determining susceptibility and resistance. Means + standard deviations of C_q values are shown for each MIC. The best-fit lines and Pearson regression R² values are shown. All drugs exhibited a statistically significant correlation, except for cycloserine.