Performance of a next-generation rapid lipoarabinomannan urine assay for tuberculosis in outpatients with and without HIV in South Africa and Uganda: a two-centre diagnostic accuracy study 

ABSTRACT

BACKGROUND: Non-sputum-based tuberculosis tests are a global priority because sputum-dependent diagnostics systematically miss populations at highest risk of disease, including people living with HIV. We evaluated the diagnostic accuracy of a next-generation urine lipoarabinomannan assay.

METHODS: In this two-centre diagnostic accuracy study, we enrolled adults aged 18 years or older, regardless of their HIV status, with presumptive pulmonary tuberculosis, attending outpatient clinics in South Africa and Uganda. Participants who had received tuberculosis treatment within the previous 2 months or had an unknown treatment history were excluded. Demographic and clinical data, including HIV status and CD4 cell count, were collected alongside urine, blood, and sputum. We compared the diagnostic accuracy of the Biopromic lipoarabinomannan assay (BP-LAM; Biopromic TB LAM, Biopromic, Solna, Sweden) and WHO-recommended Determine TB LAM Ag assay (LF-LAM; Abbott, Santa Clara, CA, USA) against an extended microbiological reference standard (eMRS; sputum Xpert MTB/RIF Ultra [Ultra; Cepheid, Sunnyvale, CA, USA], culture, or both). Sputum induction was available in South Africa. Secondary outcomes were diagnostic yield and accuracy stratified by subgroups (HIV, CD4 cell count stratum, country, and sputum scarcity status), including design-locked and prototype assays.

FINDINGS: We tested urine from 629 participants (282 from South Africa and 347 from Uganda) collected between Oct 5, 2018, and March 11, 2022. 315 individuals with tuberculosis and a random selection of 314 without tuberculosis underwent urine testing. 250 (40%) of 626 participants were female and 376 (60%) were male; the median age was 34 years (IQR 19-60). BP-LAM had higher sensitivity than LF-LAM (63% [95% CI 58-69] vs 22% [18-27]; p<0·0001) and similar specificity (93% [90-96] vs 89% [85-92]; p=0·067). BP-LAM sensitivity was similar regardless of HIV status (67% [58-76] for people with HIV vs 62% [55-69] for people who were HIV-negative; p=0·39) but was higher in people with HIV and CD4 counts less than or equal to 200 cells per μL (79% [63-90] vs 58% [44-72] for people with HIV and CD4 count >200 cells per μL; p=0·041). Specificity of BP-LAM was significantly lower in people with HIV than in individuals who were HIV-negative (90% [83-94] vs 96% [92-99]; p=0·016). The design-locked BP-LAM had better specificity than the prototype (93% [90-96] vs 85% [80-88]; p<0·0001). Diagnostic yield was significantly higher for Ultra (82% [78-86]), followed by culture (72% [68-77]) in comparison with BP-LAM (61% [56-66]; p<0·0001), and LF-LAM yield was significantly lower (29% [24-33]; p<0·0001), with similar patterns across HIV status. BP-LAM yield was similar to that of Ultra when sputum induction was unavailable (67% [55-79] vs 64% [49-78]; p=0·82).

INTERPRETATION: BP-LAM showed higher sensitivity than LF-LAM, including in HIV-negative people, and identified a similar number of cases as sputum Ultra when sputum induction was unavailable. Further validation and implementation studies are warranted.

FUNDING: Global Health Technology Fund (GHIT) programme and EDCTP2 programmes supported by the EU.

PMID:42561988 | DOI:10.1016/j.lanmic.2026.101418