Xpert Ultra on nasopharyngeal swabs detects one-third of pulmonary tuberculosis: a nested case-control study in South African primary care

Background Sputum-free alternatives for tuberculosis (TB) diagnosis are a public health priority. Nasopharyngeal swabs (NPSs), a specimen routinely used for viral testing especially at-scale in outbreak investigations, may have promise but has not been evaluated in people investigated for TB. Methods We evaluated Xpert MTB/RIF Ultra (Ultra; Cepheid, Sunnyvale, California, USA) on stored NPSs (NPSU) from adults with presumptive TB consecutively presenting to primary care in Cape Town, South Africa. Using a nested case-control design, all participants with TB had a NPS tested (n = 148) and people without TB (n = 117) were randomly selected matched by HIV status. We used a microbiological (MRS) and extended microbiological reference standards (eMRS). A subset of convenience-sampled cases-controls also received flocked tongue swab Ultra (TSU) (n = 57 cases, n = 66 controls). Results Twenty-one (8%) of NPSUs were unsuccessful (all errors). NPSU sensitivity was 35% (95% CI 27, 43), specificity 88% (80, 93). Sensitivity was lower in HIV-positive participants [24% (14, 36) vs. 44% (32, 56)]. TSU had higher sensitivity [67% (53, 79) vs. 28% (17, 42) in the head-to-head comparison, detecting 22 additional cases] than NPSU. Specificity was comparable. Conclusion NPSU sensitivity (35%) was lower than TSU and below the WHO target of 80% for non-sputum confirmatory tests; specificity (88%) was also below the 98% target. NPSs may have utility if already collected in mass testing campaigns for other diseases but unless performance can be optimized (e.g., by using purpose-built swab platforms) do not appear useful for TB diagnosis.