The epidemiology of tuberculosis in urban populations is changing. Combining conventional epidemiologic techniques with DNA fingerprinting of Mycobacterium tuberculosis can improve the understanding of how tuberculosis is transmitted.

We used restriction-fragment-length polymorphism (RFLP) analysis to study M. tuberculosis isolates from all patients reported to the tuberculosis registry in San Francisco during 1991 and 1992. These results were interpreted along with clinical, demographic, and epidemiologic data. Patients infected with the same strains were identified according to their RFLP patterns, and patients with identical patterns were grouped in clusters. Risk factors for being in a cluster were analyzed.

Of 473 patients studied, 191 appeared to have active tuberculosis as a result of recent infection. Tracing of patients' contacts with the use of conventional methods identified links among only 10 percent of these patients. DNA fingerprinting, however, identified 44 clusters, 20 of which consisted of only 2 persons and the largest of which consisted of 30 persons. In patients under 60 years of age, Hispanic ethnicity (odds ratio, 3.3; P = 0.02), black race (odds ratio, 2.3; P = 0.02), birth in the United States (odds ratio, 5.8; P<0.001), and a diagnosis of the acquired immunodeficiency syndrome (odds ratio, 1.8; P = 0.04) were independently associated with being in a cluster. Further study of patients in clusters confirmed that poorly compliant patients with infectious tuberculosis have a substantial adverse effect on the control of this disease.

Despite an efficient tuberculosis-control program, nearly a third of new cases of tuberculosis in San Francisco are the result of recent infection. Few of these instances of transmission are identified by conventional contact tracing. .