BACKGROUND: The DOTS (directly observed treatment short-course) strategy for tuberculosis (TB) control is recommended by the World Health Organization globally. However, there are few studies of long-term TB treatment outcomes from DOTS programs in high-burden settings and particularly settings of high drug resistance. A DOTS program was implemented progressively in Karakalpakstan, Uzbekistan starting in 1998. The total case notification rate in 2003 was 462/100,000, and a drug resistance survey found multidrug-resistant (MDR) Mycobacterium tuberculosis strains among 13% of new and 40% of previously treated patients. A retrospective, observational study was conducted to assess the capacity of standardized short-course chemotherapy to effectively cure patients with TB in this setting. METHODS AND FINDINGS: Using routine data sources, 213 patients who were sputum smear-positive for TB, included in the drug resistance survey and diagnosed consecutively in 2001-2002 from four districts, were followed up to a median of 22 months from diagnosis, to determine mortality and subsequent TB rediagnosis. Valid follow-up data were obtained for 197 (92%) of these patients. Mortality was high, with an average of 15% (95% confidence interval, 11% to 19%) dying per year after diagnosis (6% of 73 pansusceptible cases and 43% of 55 MDR TB cases also died per year). While 73 (74%) of the 99 new cases were "successfully" treated, 25 (34%) of these patients were subsequently rediagnosed with recurrent TB (13 were smear-positive on rediagnosis). Recurrence ranged from ten (23%) of 43 new, pansusceptible cases to six (60%) of ten previously treated MDR TB cases. MDR M. tuberculosis infection and previous TB treatment predicted unsuccessful DOTS treatment, while initial drug resistance contributed substantially to both mortality and disease recurrence after successful DOTS treatment. CONCLUSIONS: These results suggest that specific treatment of drug-resistant TB is needed in similar settings of high drug resistance. High disease recurrence after successful treatment, even for drug-susceptible cases, suggests that at least in this setting, end-of-treatment outcomes may not reflect the longer-term status of patients, with consequent negative impacts for patients and for TB control.