Recent outbreaks of multi-drug-resistant tuberculosis (MDR-TB) have resulted in significant morbidity and mortality in patients with AIDS. The poor outcomes are attributable to delayed diagnoses, slow reporting of antimycobacterial susceptibility results, inadequate treatment regimens and profound immunosuppression. There are no prospective clinical trials which have evaluated the optimal treatment of MDR-TB. A retrospective study has shown that in immunocompetent patients with secondary MDR-TB, only 56% responded to prolonged courses of multiple drug regimens, and 22% died of TB. In patients with AIDS, even fewer patients respond, with median survivals of 2-4 months. In general, better responses have been associated with in vitro susceptibility of patients' isolates. If possible, patients with MDR-TB should receive at least three drugs to which their isolates are susceptible for at least 24 months; these regimens are likely to include ethambutol, pyrazinamide, a quinolone, and an aminoglycoside. Selected patients benefit from surgical intervention combined with aggressive chemotherapy. MDR-TB is best prevented by directly observed therapy of patients with susceptible organisms and rigorous infection control practices in areas of high incidence of MDR-TB. Effective treatment regimens for MDR-TB await the development of novel compounds which have better in vitro activity against MDR-TB than currently available drugs.