A gold mining work force was followed prospectively over 1 yr for sputum nontuberculous mycobacterial (NTM) isolates. NTM were isolated from 118 men, of whom 32 (27%) met the American Thoracic Society (ATS) case-definitions for pulmonary NTM disease (23 M. kansasii, seven M. scrofulaceum, one M. avium, and one M. abscessus). Determining isolate significance was difficult because most men had been started on presumptive TB treatment before isolate identification (70%). Histologic criteria were considered inappropriate for this high M. tuberculosis incidence population, particularly for patients who had stabilized on presumptive TB treatment. Among men not meeting case-definitions, indicators of disease were significantly more prevalent for M. kansasii than for M. fortuitum, the local laboratory contaminant (ORs: 6.5 for cough, 7. 2 for smear-positivity, 36.0 for radiologic changes, and 14.3 for presumptive TB treatment), suggesting underdiagnosis of M. kansasii disease. Of 53 men with definite or possible M. kansasii disease, 18 (34%) were HIV-positive. HIV-associated M. kansasii disease occurred at an early stage of immunosuppression (median CD4 count, 381 x 10(6)/L) with a good outcome (83% cured after 12 mo of treatment). ATS case-definitions for NTM disease are difficult to apply in this population, and treatment decisions should be guided by the pathogenic potential of the isolate.