Testing tongue swab samples by Cepheid Xpert MTB/RIF Ultra: Comparison of two protocols applied to samples from persons with low-bacillary load TB

Tongue swabs (TS) are novel non-sputum specimens for molecular diagnosis of tuberculosis (TB). Analysis of swabs by using Cepheid Xpert MTB/RIF Ultra® (Xpert Ultra) is less sensitive than the use of certain manual qPCR methods but is desirable given the widespread use and familiarity of Xpert Ultra. This study evaluated an easy-to-use protocol for analyzing TS by Xpert Ultra. TS samples from symptomatic South African patients were previously collected and tested by Xpert Ultra using a protocol that requires a heating block (termed Heat+TE). Replicate, paired samples from the same patients were tested by a newly reported method that does not require heating (termed 2:1 SR). The diagnostic accuracy of the two methods was compared. Because the sensitivity of TS testing is lowest in patients with low sputum bacillary loads, the resolving power of this paired comparison was maximized by prioritizing samples from such participants (N=108), 88 of whom were TB-positive by sputum microbiological reference standard (MRS) and 20 were TB-negative. Within this sample set, the Heat+TE method was 42.5% (38/88; 95CI 33-54%) sensitive and the 2:1 SR method was 52.3% (46/88; 95CI 41-63%) sensitive relative to sputum MRS. Both methods were 100% (20/20; 95% CI 83-100%) specific. A secondary analysis compared the sensitivities of single-sample versus dual-sample testing within a cohort (N=241) that included both high- and low-bacillary load participants. Testing of a single sample was 75.5% (182/241; 95CI 70-81%) sensitive relative to sputum MRS, whereas dual sampling was 81.3% (196/241; 95CI 76-86%) sensitive.