Résumé
Tuberculosis (TB) remains under-diagnosed in people living with HIV (PLHIV) and is a major cause of mortality in these population. Sputum is the primary specimen for TB diagnosis; however, many PLHIV are unable to produce quality sputum. This study assessed the diagnostic utility of alternative non-invasive specimens, stool, and urine, for TB detection in PLHIV using the Xpert MTB/RIF Ultra assay.
A cross-sectional study was conducted in TB screening and treatment centres in Conakry, Guinea, including 300 PLHIV presumptive pulmonary TB. Sputum, stool, and urine specimens were collected and tested using the Xpert MTB/RIF Ultra assay. Diagnostic performance was assessed in terms of sensitivity, specificity, and positive and negative predictive values, with their 95% confidence intervals, using the sputum Xpert MTB/RIF Ultra results on as the reference standard. Agreement between specimen results was assessed using Cohen's kappa.
Of the 300 PLHIV included in the study, 52% (156/300) were male. Median age was 37 years (IQR: 30-45), with more than two-thirds under 45 years of age. Most patients were newly diagnosed with HIV in 2024 (93.0%; 279/300), and 80.0% (240/300) were not yet on antiretroviral therapy. TB positivity was 22.3% (67/300) in sputum, 24.7% (74/300) in stool, and 20% (60/300) in urine. The Xpert MTB/RIF Ultra assay demonstrated a sensitivity of 86.5% and a specificity of 93.1% in stool specimens, compared to 56.7% and 90.5% in urine specimens. Sensitivity was highest in patients with CD4 < 200 cells/mm
, 89.4% (42/47) for stool and 61.7% (29/47) for urine, while specificity remained consistently high across all CD4 categories, reaching 100% in patients with CD4 counts ≥ 500 cells/mm
. Substantial agreement was observed between stool and sputum results (kappa = 0.77; p < 0.001).
Xpert MTB/RIF Ultra assay demonstrated high diagnostic performance on stool specimens in PLHIV, with sensitivity surpassing that of urine specimens. Stool testing could serve as a complementary diagnostic approach, particularly for patients unable to produce sputum, thereby improving TB detection in PLHIV. Integrating stool testing into national TB diagnostic algorithms could enhance case detection and management in PLHIV, although implementation must be assessed for technical feasibility and cost-effectiveness.
Not applicable.