Infections with >1 Mycobacterium tuberculosis strain(s) are underrecognized. We show, in vitro and in vivo, how first-line treatment conferred a competitive growth advantage to amplify a multidrug-resistant M. tuberculosis strain in a patient with mixed infection. Diagnostic techniques that identify mixed tubercle bacilli populations are needed to curb the spread of multidrug resistance.A s the number of multidrug-resistant tuberculosis (TB) cases continues to rise, so does the amplification of multidrug-resistant Mycobacterium tuberculosis strains during treatment (1,2). This amplification is generally assumed to result from in vivo evolution of drug resistance caused by poor therapy compliance or, in high-incidence settings, from exogenous reinfection with a multidrug-resistant strain. We report a case in which emergence of multidrug resistance did not result from in vivo acquisition of drug resistance by a drugsensitive strain or from exogenous reinfection with an already resistant strain. By integrating epidemiologic, microbiological, and molecular strain typing data with in vitro competitive growth experiments, we provide evidence for an initial mixed infection with a drug-sensitive strain and an undetected drug-resistant strain that outgrew the sensitive strain under the selection pressure of first-line chemotherapy.M. tuberculosis strains in sputum from TB-infected patients or in samples from the disease site are generally identified by strain typing a single broth culture or colony grown on solid medium. However, this method does not enable identification of mixed infections, and any treatment regimen would be determined on the basis of the drug sensitivity of the strain with the fastest growth rate in the in vitro culture. Use of suboptimal drug combinations could lead to selection of a slower growing, drug-resistant strain already present in the host and thus to treatment failure.Studies of artificially mixed M. tuberculosis strains before and after culture showed that culturing can reduce the clonal complexity of the strains and that, in most samples (6/10), only 1 strain will be identified in mixed infections after culture (3). This suggests that mixed infections and clonal complexity are underrepresented in culture-based diagnoses of TB. In support of this suggestion, the results of molecular-based methods that use strain-specific PCR showed that 2.1%-19.0% of patients with active TB in moderate to high incidence countries were simultaneously infected with >2 strains (1,2,4-10).Possible co-infection of patients with drug-sensitive and drug-resistant M. tuberculosis strains has been described (1,2), and modeling of the effect of such co-infection on the long-term dynamics of tuberculous infection has led to the hypothesis that persons with this type of infection may retain small populations of drug-resistant bacteria that can flourish after the host receives treatment (11). van Rie et al. showed the amplification of a drug-resistant strain after treatment and postulated selection of drug-resistant st...