Comparison of Nucleic Acid Amplification Tests for Tuberculosis

Piersimoni et al. have reported their experience in parallel testing of the Amplified Mycobacterium tuberculosis Direct Test (AMTDII) (Gen-Probe Inc., San Diego, Calif.) and the Abbott LCx Mycobacterium tuberculosis Assay (LCx) (Abbott Laboratories Diagnostic Division, Abbott Park, Ill.) with 273 respiratory samples and 184 nonrespiratory samples (1). Their study provides further useful statistics on the value of the AMTDII and LCx as diagnostic tools for tuberculosis (TB).

In general, the results presented by Piersimoni et al. support the contention that regardless of the manufacturer, commercial amplification tests have almost absolute sensitivity (and specificity) when applied to microscopy-positive acid-fastbacillus (M-afb) samples. While results with M-afb samples will be of undeniable value—particularly in localities with significant rates of disease due to atypical mycobacteria—the preferred commercial test is likely to be the one that performs best with (i) samples that are microscopy negative (M-neg) but which grow M. tuberculosis on culture (C-MTB) and (ii) samples from patients who are known to have TB but which are M-neg and negative by culture (C-neg). Numerous earlier studies, using different assays and a variety of samples, have suggested widely divergent sensitivities with such material.

Piersimoni et al. acknowledge the need for parallel testing with various kits, and their study was designed to address this issue. Nevertheless, the authors seem to have missed the opportunity to present some valuable comparative data. Before looking at specific examples, it should be noted that the abstract of their article states that “the level of agreement between AMTDII and LCx assay results was 78.2%.” But there is no reference to this statistic in the body of the paper, and the data (as presented) do not allow the reader to check this figure. The study found that AMTDII “missed” 4 (10.5%) of 38 samples that were M-neg and C-MTB, whereas LCx missed 17 (44.7%). Further, in testing 25 samples which were from patients with TB but which were M-neg and C-neg, both assays missed 11 (44.0%). These findings beg the question, Were the samples missed by AMTDII also missed by LCx? The authors’ conclusion that AMTDII is significantly more sensitive than LCx with both respiratory and extrapulmonary samples is of particular interest to laboratories looking for guidance in choosing a commercial assay. However, a table showing correlated results for AMTDII and LCx with the subset of samples that gave inconsistent results would have been most valuable in assessing the performance of the individual assays.