Performance of three diagnostic tests for tuberculosis among Human Immunodeficiency Virus (HIV) patients presenting to a health facility in an informal urban settlement in Nairobi, Kenya.

The absence of an accurate reference test complicates the evaluation of tuberculosis (TB) diagnostic tests among people living with Human Immunodeficiency Virus (PLWHIV). The objective of this study was to estimate (using Bayesian latent class models [BLCM]) the sensitivity (Se), specificity (Sp) and negative and positive predictive values (NPV and PPV) of sputum smear microscopy (SSM), Xpert Ultra and lipoarabinomannan antigen (LAM) tests for TB among PLWHIV in Nairobi, Kenya. This cross-sectional study enrolled a total of 190 patients aged ≥ 18 years with presumptive TB seeking treatment at the Kibra Community Health Center Comprehensive Care Centre (CCC) clinic between September 2022 and March 2023. The diagnostic data obtained from the three tests were analysed using a BLCM framework to derive accuracy estimates of the three diagnostic tests. The Xpert Ultra assay registered a higher Se (85.0; 95% PCI [41.4-99.4]) compared to LAM (26.8; 95% PCI [4.7-67.6]) and SSM (56.7 [16.4-97.4]). However, SSM had the highest Sp (99.6; 95% PCI [97.7-100.0]). The Xpert Ultra assay yielded the highest overall combination of Se and Sp at 80.8% (95% PCI [37.0-96.5]). On predictive values, SSM recorded the highest PPV at 84.5% (95% PCI [38.4-99.4]). Nonetheless, all the tests exhibited noticeably high NPVs (>96%). The Xpert Ultra assay recorded the highest Se of the three tests. Nevertheless, the SSM test registered the highest Sp and correspondingly the highest PPV. On NPVs, the tests had similar estimates. Consequently, a two-test serial testing strategy, entailing an initial screening with the more sensitive Xpert Ultra assay followed by retesting of any positives with the more specific SSM test, could be most optimal in this low-burden TB setting.