SETTING: Public ambulatory centers in northern Lima, Peru. OBJECTIVE: To compare two retreatment strategies in Category I failures. DESIGN: Retrospective cohort study of Category I failures enrolled between February 1997 and October 2001. Strategy A was a nationwide approach, applying a Category II regimen; if that regimen failed, a standardized regimen including second-line drugs was used. Strategy B was a pilot protocol designed to diagnose and treat multidrug-resistant tuberculosis (MDR-TB); this strategy included drug susceptibility testing (DST) and eliminated the Category II regimen. RESULTS: Of 125 patients that Category I failed to cure, 73 entered Strategy A and 52 entered Strategy B. Almost 90% of those with DST results had MDR-TB. Strategy B was three times more likely than Strategy A to cure patients (79% vs. 38%, RR = 2.9, 95% CI 1.7-5.1) and five times more likely to cure patients than the Category II regimen alone (79% vs. 15%, RR 5.2, 95%CI 3.0-9.2). Strategy B also significantly reduced delays to MDR-TB diagnosis and to the initiation of MDR-TB therapy. CONCLUSIONS: Under program conditions, a retreatment strategy based on DST and eliminating the Category II regimen can improve clinical outcomes among Category I treatment failures found to have active, infectious MDR-TB. © 2005 The Union.