Abstract
Numerous tuberculosis (TB) deaths remain undetected in low-resource endemic settings. With autopsy-confirmed tuberculosis as our standard, we assessed the diagnostic performance of Xpert MTB/RIF Ultra (Ultra; Cepheid) on nasopharyngeal specimens collected postmortem.
From October 2016 through May 2019, we enrolled pediatric and adult medical deaths to a prospective autopsy study at two referral hospitals in northern Tanzania with next-of-kin authorization. We swabbed the posterior nasopharynx prior to autopsy and tested the samples later by Ultra. At autopsy we collected lung, liver, and, when possible, cerebrospinal fluid for mycobacterial culture and histopathology. Confirmed tuberculosis was defined as Mycobacterium tuberculosis complex recovery by culture with consistent tissue histopathology findings; decedents with only histopathology findings, including acid-fast staining or immunohistochemistry, were defined as probable tuberculosis.
Of 205 decedents, 78 (38.0%) were female and median (range) age was 45 (0,96) years. Twenty-seven (13.2%) were found to have tuberculosis at autopsy, 22 (81.5%) confirmed and 5 (18.5%) probable. Ultra detected M. tuberculosis complex from the nasopharynx in 21 (77.8%) of 27 TB cases (sensitivity 70.4% [95% confidence interval {CI} 49.8–86.2%], specificity 98.9% [95% CI 96.0–99.9%]). Among confirmed TB, the sensitivity increased to 81.8% (95% CI 59.7–94.8%). Tuberculosis was not included as a death certificate diagnosis in 14 (66.7%) of the 21 MTBc detections by Ultra.
Nasopharyngeal Ultra was highly specific for identifying in-hospital tuberculosis deaths, including unsuspected tuberculosis deaths. This approach may improve tuberculosis death enumeration in high-burden countries.