cepted as a measure of the risk of tuberculosis in dually infected persons (94). He found a tuberculosis risk of 7.9/100 person-years in PPD-positive, HIV-infected addicts and a risk of 0.3/100 person-years in PPD-negative, HIV-seropositive individuals. Isoniazid treatment of latent tuberculous infection was offered to all tuberculin-positive addicts, but the rate of compliance was low.

In a second study in 1987–1990, Selwyn and collaborators documented patient compliance with isoniazid treatment and also used a battery of delayed hypersensitivity antigens to assess skin test anergy (95). Among 25 persons who did not complete isoniazid, the tuberculosis incidence was 9.7/100 person-years. No cases occurred among those who completed treatment of latent infection. Among anergic individuals, the incidence was 6.6/100 person-years.

Useful information on the risk of tuberculosis in HIV-infected persons exposed to tuberculosis comes from two outbreak studies. Di Perri and coworkers described an outbreak of tuberculosis among AIDS patients in a hospital in Italy at which no patient with tuberculosis had been hospitalized during the previous 3 years (96). Ventilation included air recirculation. An individual with initially unsuspected tuberculosis was admitted, and an outbreak ensued. Among 18 exposed HIV-infected individuals, tuberculosis developed in eight, seven of them within 60 days.

Daley and colleagues described an outbreak in a congregate living facility for AIDS patients in San Francisco (97). A person with unrecognized tuberculosis was admitted to the facility; the diagnosis was made 6 weeks later after 3 weeks of progressive respiratory symptoms. Eleven of 30 exposed residents developed tuberculosis within the next 6 months, and organisms isolated from them were all of the same RFLP type as the organism from the index case. These two outbreaks demonstrate the enormous impact of HIV infection on susceptibility to tuberculosis.

In sum, one should probably accept a 10 percent annual risk of disease among tuberculin-positive persons who become HIV infected and a 35 to 45 percent early disease risk among AIDS patients who acquire infection with M. tuberculosis. Appropriate treatment of latent infection in both groups should reduce these risks.

Clinical trials of antituberculosis therapeutic regimens conducted in the United States and elsewhere beginning the 1950s demonstrated low mortality rates among adequately treated individuals. The only modern American data come from United States Public Health Service trial number 21, a multicentered national trial of modern drug treatment regimens. Nine deaths occurred among 1,451 participants, 0.6 percent (98). HIV