OBJECTIVES: An accurate test for Mycobacterium tuberculosis infection is urgently needed. The tuberculin skin test (TST) lacks sensitivity, particularly in HIV-infected individuals, and has poor specificity because of antigenic cross-reactivity with Bacillus Calmette-Guérin (BCG) vaccination. ESAT-6 and CFP-10 are antigens expressed in Mycobacterium tuberculosis, but not in Mycobacterium bovis BCG and most environmental mycobacteria. We investigated whether T cells specific for these antigens could serve as accurate markers of M. tuberculosis infection in an area of high tuberculosis and HIV prevalence. METHODS: Using the rapid ex-vivo enzyme-linked immunospot (ELISPOT) assay for IFN-gamma, we enumerated T cells specific for ESAT-6, CFP-10 and purified protein derivative (PPD) in blood samples from 50 Zambian tuberculosis patients, 75 healthy Zambian adults, and 40 healthy UK residents. TSTs were performed in 49 healthy Zambian adults. RESULTS: All (100%; n = 11) and 90% (n = 39) of HIV-negative and HIV-positive tuberculosis patients, respectively, had detectable ESAT-6- or CFP-10-specific T cells. The ESAT-6/CFP-10-based ELISPOT assay was positive in 37 out of 54 HIV-negative healthy Zambians, suggesting a 69% prevalence of latent M. tuberculosis infection. Fewer HIV-positive Zambians possessed ESAT-6/CFP-10-specific T cells, but the impact of HIV infection was less on this assay than on the PPD-based ELISPOT or TST. CONCLUSION: The ESAT-6/CFP-10-based ELISPOT assay detects active tuberculosis in HIV-positive individuals with high sensitivity. It is more specific, and possibly more sensitive, than PPD-based methods of detecting latent M. tuberculosis infection, and may potentially improve the targeting of isoniazid preventative therapy to HIV-positive individuals with latent tuberculosis infection.