RR:C19 Evidence Scale rating by reviewer:
Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.
The authors advance their work on TB molecular testing of tongue swabs among ambulatory South Africans with TB symptoms with 3 principal findings: 1) use of a dry preparation of COPAN tongue swabs for TB molecular testing was similarly sensitive and specific as in a previous study that used a wet swab preparation involving collection and storage in buffer; 2) a streamlined 15-minute protocol for processing tongue swabs for GeneXpert MTB/RIF Ultra performed similarly to previous studies that used a 45-60 minute processing protocol; and 3) swab testing with MolBio MTB Ultima had similar sensitivity but lower specificity than MTB Xpert Ultra.
Wood and colleagues continue to advance their work on the feasibility and diagnostic performance of tongue swabs for molecular diagnosis of TB in the largest known study to date, which included 321 South Africans with TB symptoms attending TB clinics in Worcester, in the Western Cape Province. The authors' had three principal findings. First, they report that nylon-flocked swabs (COPAN Diagnostics) that were collected and stored dry gave similar sensitivity and specificity as found in previous studies that used wet swabs collected and stored in buffer. Second, they found that an expedited 15-minute swab processing protocol provided similar diagnostic accuracy as previous studies that used a 45-60 minute protocol. Third, they report that testing swabs using the MolBio MTB Ultima assay provided similar sensitivity but lower specificity than GeneXpert Ultra in identical and overlapping population subsets.
An important secondary finding was that retesting of false negative MolBio MTB Ultimate swabs had a modest impact, increasing sensitivity by only about 7.5%. Since the testing protocol did not routinely including internal positive controls, inhibition may be another possible explanation for this particular study finding.
This study has several important strengths, including its rigorous sample processing and testing protocols and systematic assessment of the impact of different sample collection, processing, and testing protocols on diagnostic accuracy. The retesting of all false negative results and a random sample of true negative results was another rigorous feature.
This study also has a few limitations. First, the collection and processing protocol comparisons both use external populations rather than internal controls, raising the possibility that the similar results are due not to equally valid protocols but rather due to population differences (e.g. an inferior protocol testing in a more severely ill population). Second, the use of a convenience rather than a consecutive sample of participants as used in this study has previously been associated with upwardly biased estimates of diagnostic test sensitivity and specificity in systematic reviews. The limited clinical information on disease burden or severity (e.g. symptom duration, chest radiography findings) is a barrier to comparing the population in this study to that of previous studies as a way of understanding the possible person-related sources of heterogeneity. The population subset for whom Xpert semi-quantitative data is reported suggests that this study enrolled a primarily pluri-bacillary population (because three-quarters had Xpert results of medium or high) if this subset is representative of the whole. Finally, it seems that 9 Xpert and 34 MTB Ultima samples were excluded from the diagnostic accuracy reporting, although it is usually recommended only by including error and invalid results that we can fairly assess the real-world diagnostic performance. This imbalance in invalid results would seem to substantially favor the performance of Xpert Ultra over MTB Ultima.
Finally, I was intrigued by the author's note that they included a heat-inactivation step for occupational safety. This is a very important implementation consideration for resource-constrained settings, and it would be exciting to hear how effective heat inactivation and/or other alternatives are for this purpose.