Use of receiver operating characteristic curves to assess the performance of a microdilution assay for determination of drug susceptibility of clinical isolates of Mycobacterium tuberculosis.

The aim of this study was to apply receiver operating characteristic (ROC) analysis to the microplate Alamar blue assay, a recently developed alternative for drug susceptibility testing of mycobacteria. As this is a quantitative assay, its performance can be determined by ROC analysis, in which the area under the ROC curve represents a summary of test performance (the higher the area, the better the test's performance). Sixty isolates of Mycobacterium tuberculosis were tested by the microcolorimetric assay against six twofold dilutions of streptomycin, isoniazid, rifampin, and ethambutol. For each isolate, the susceptibility pattern was simultaneously established by the agar proportion method, the result of which represented the gold standard value for the ROC analysis. The critical concentration, area under the curve, and P value for each drug were determined by ROC curve analysis. The results of the assay were obtained in an average of 8 days of incubation. The performance of the assay was excellent for all four drugs: the area under the curves was >0.97, the P values were 0.000, and sensitivity was 94%, specificity 97%, predictive value for resistance >/=92%, predictive value for susceptibility 97%, and test efficiency 97%. According to ROC analysis, the microplate Alamar blue assay is a reliable method for determination of drug-susceptibility. Rapidity and cost efficiency are two additional qualities that make this test an excellent alternative for the drug susceptibility testing of Mycobacterium tuberculosis. The ROC curve analysis is a robust statistical approach for evaluating the performance of new quantitative methods for determination of drug sensitivity of Mycobacterium tuberculosis isolates.