From September 1996 to June 1997, in an area bordering East London, we prospectively collected epidemiological, clinical and microbiological data on all patients with newly diagnosed culture-positive tuberculosis and compared these to national data based on notifications. The significant differences were that tuberculosis was diagnosed almost exclusively in non-Caucasian patients (42/47 [89%]) and that there was a high percentage of extrapulmonary tuberculosis (27/47 [57%]) including four cases of tuberculous meningitis and five cases of osteomyelitis. We also observed that 19/27 (70%) of patients with extrapulmonary tuberculosis had normal chest X-rays, 3/17 (18%) sub-Saharan Africans were HIV antibody-positive and drug resistance strains were isolated from six sub-Saharan Africans and one Caucasian. Figures for treatment failures and mortality compared favorably to national averages at 6 months. National data do not accurately reflect local epidemiology and clinical presentations. Hospital-based surveillance and promoting awareness of local differences is essential to prevent delayed diagnosis, inappropriate management and poor clinical outcome.