Setting: A DOTS clinic in an academic tertiary referral hospital in South India.
Objective: To evaluate the performance of two rapid enzyme-linked immunoassays (EIAs) for the detection of human immunodeficiency virus (HIV) infection in sputum samples of patients with tuberculosis (TB).
Design: We prospectively recruited 522 consecutive out-patients presenting to the DOTS clinic with confirmed TB of any type to undergo HIV testing using reference serum EIA and index-blinded parallel sputum HIV testing with two rapid EIAs designed for oral mucosal transudate.
Results: HIV positivity was 14.9% (95%CI 12.1-18.4). Compared to reference serum EIA, the Oraquick™ assay was 93.1% sensitive (95%CI 83.8-97.4) and 95.3% specific (95%CI 92.7-96.9), while the Aware™ assay was 92.3% sensitive (95%CI 83.4-96.8) and 96.6% specific (95%CI 94.4-98.0). The positive predictive values were respectively 77.0% and 82.7%. After freezing of sputum, the sensitivity of both assays declined, but the specificity significantly increased. Higher sputum volume reduced the odds of obtaining a true result with both assays.
Conclusion: HIV testing of fresh sputum is not sufficiently accurate for anonymous HIV surveillance among TB patients in a setting of low (<10%) HIV prevalence. Freezing sputum samples and limiting sputum volume for HIV testing may improve assay specificity.