Anergy testing has been used as an adjunct to tuberculin testing for assessing M. tuberculosis (MTB) infection and indications for isoniazid preventive therapy in HIV-infected persons. We examined factors associated with the stability of skin test responses to purified protein derivative (PPD) and candida antigens in a cohort of HIV-infected adults followed prospectively in a tuberculosis preventive therapy trial in Uganda. PPD-positive and anergic subjects in the placebo arms of the preventive therapy study underwent repeat skin testing and immunologic testing including measurement of MTB culture filtrate (CF)-stimulated interferon gamma (IFN-gamma) and tumor necrosis factor alpha (TNF-alpha) levels in whole-blood culture supernatants. Anergy was present in 27% of 4,058 HIV-infected subjects screened for the tuberculosis preventive therapy trial compared with 10% of 682 HIV-non-infected persons. On follow-up testing of enrolled subjects, 42% of 139 initially anergic subjects were no longer anergic; two thirds of these had PPD reactions >= 5 mm. Stability of anergy was associated with intercurrent opportunistic infections and AIDS-associated dermatitis at baseline. Thirty-five percent of 313 subjects with an initial positive PPD had a negative PPD test at follow-up, 26% of whom had a positive candida skin test at the same time as the negative PPD test. Baseline MTBCF-stimulated IFN-gamma levels were significantly higher among PPD-positive subjects who remained PPD-positive than in those who were falsely negative. We conclude first that anergy is unstable and second that anergy testing is unreliable in identifying HIV-infected adults who are not infected with MTB and should not be used routinely for this purpose in assessing indications for isoniazid preventive therapy.