Outcome 4-6 years after retreatment was assessed for Phases 3 and 4 of the Toronto Study. Of 477 teeth retreated, 333 were lost to follow-up, 18 were extracted, and 126 (41% recall, excluding 124 discontinuers) were examined for outcome of healed (periapical index score, < or = 2; no signs or symptoms) or diseased. When pooled with Phases 1 and 2, 187 of 229 teeth (82%) were healed. Logistic regression identified significant (P < or = .05) preoperative outcome predictors: root filling quality (odds ratio [OR], 4.18; confidence interval [CI], 1.72-10.12; healed: inadequate, 88%; adequate, 66%), perforation (OR, 4.01; CI, 1.28-12.62; healed: absent, 87%; present, 56%), and radiolucency (OR, 3.33; CI, 1.19-9.36; healed: absent, 93%; present, 80%). In teeth with radiolucency, outcome predictors were number of treatment sessions (OR, 12.08; CI, 1.84-infinity; healed: one, 100%; > or = 2, 77%) and previous root filling quality (OR, 7.68; CI, 2.36-26.89; healed: inadequate, 86%; adequate, 50%). Outcome was better in teeth with inadequate previous root filling, without perforation and radiolucency. Suggested significance of number of treatment sessions in teeth with radiolucency requires validation from randomized controlled trials.