Objective: Surgical time prediction is an important factor to plan both clinical and organizational aspects of mandibular impacted third molar extraction. Many classifications have been proposed over the years, but their accuracy in surgical time prediction remained questionable. The present study introduced a modification of Juodzbalys and Daugela (JD) classification, and had the aim to validate its effectiveness in predicting the duration of the surgery. Method and Materials: Three centers treated patients needing impacted mandibular third molar extraction, following inclusion and exclusion criteria. Extractions were performed following a standardized approach, and surgical time was recorded. A blinded assessor assigned scores to each extracted tooth, according to original and modified JD classifications. Differences among the operators were evaluated though Kruskal-Wallis test, and backward multiple linear regressions were performed to evaluate the variables associated with surgical time, considered as the main outcome of the study. Results: 124 patients were treated with mandibular third molar extraction. Mean surgical time was 24.1 ± 22.2 minutes, with significant differences among the centers (P = .001). Surgical times among groups derived from both former and modified JD classifications were significantly different (P = .002 and P = .001, respectively). In the multivariate analysis, the statistical model including modified JD score was more efficient than the model with former JD score in predicting surgical time (R2 = .204 and R2 = .126, respectively). Conclusion: Modified JD classification resulted in a reliable tool for predicting surgical time of impacted mandibular third molar extraction; this could represent an adjunctive tool for clinician and patient in the decision-making process. Palabras clave: diagnostic procedure, extraction, oral surgery, oral surgical procedures, radiography