© 2015 Published by Oxford University Press for the Infectious Diseases Society of America 2015. Background. Resistance to second-line drugs develops during treatment of multidrug-resistant (MDR) tuberculosis, but the impact on treatment outcome has not been determined. Methods. Patients with MDR tuberculosis starting second-line drug treatment were enrolled in a prospective cohort study. Sputum cultures were analyzed at a central reference laboratory. We compared subjects with successful and poor treatment outcomes in terms of (1) initial and acquired resistance to fluoroquinolones and second-line injectable drugs (SLIs) and (2) treatment regimens. Results. Of 1244 patients with MDR tuberculosis, 973 (78.2%) had known outcomes and 232 (18.6%) were lost to follow-up. Among those with known outcomes, treatment succeeded in 85.8% with plain MDR tuberculosis, 69.7% with initial resistance to either a fluoroquinolone or an SLI, 37.5% with acquired resistance to a fluoroquinolone or SLI, 29.3% with initial and 13.0% with acquired extensively drug-resistant tuberculosis (P <. 001 for trend). In contrast, among those with known outcomes, treatment success increased stepwise from 41.6% to 92.3% as the number of drugs proven effective increased from ≤1 to ≥5 (P <. 001 for trend), while acquired drug resistance decreased from 12% to 16% range, depending on the drug, down to 0%-2% (P <. 001 for trend). In multivariable analysis, the adjusted odds of treatment success decreased 0.62-fold (95% confidence interval,. 56-.69) for each increment in drug resistance and increased 2.1-fold (1.40-3.18) for each additional effective drug, controlling for differences between programs and patients. Specific treatment, patient, and program variables were also associated with treatment outcome. Conclusions. Increasing drug resistance was associated in a logical stepwise manner with poor treatment outcomes. Acquired resistance was worse than initial resistance to the same drugs. Increasing numbers of effective drugs, specific drugs, and specific program characteristics were associated with better outcomes and less acquired resistance.