The objective of this in vitro study was to evaluate gingival microleakage in Class II total bond resin restorations in comparison to open sandwich technique restorations using different materials. Forty-eight human molar teeth were disinfected and stored in a 0.9% saline solution. In each tooth, two standardized Class II cavities (3 mm x 6 mm x 2 mm) were prepared with the gingival cavosurface margins located 1 mm below the cementoenamel junction. The teeth were divided into 4 equally sized groups (n = 12), and the proximal boxes were treated as follows: in Group 1, no base material was used and the cavity was restored using Syntac Sprint (SS) and Tetric Ceram (TC); in Groups 2, 3, and 4 the gingival portion of the cavity was restored with different base materials (Group 2, Dyract; Group 3, Vitremer; Group 4, Chelon-fil) prior to the placement of the composite resin. After a storage time of 7 days, the restorations were finished and polished. Then, specimens were submitted to thermocycling (500 cycles, 5°C to 55°C, 15 s dwell time) and immersed in a 0.5% methylene blue solution for 24 h. After washing, they were sectioned in a mesio-distal direction. Each restoration was evaluated under a stereomicroscope at 20X by 2 examiners and scored on a 0 to 3 scale according to the marginal leakage. Kappa statistics were used to evaluate the agreement between the examiners. Given the ordinal nature of the scoring system, data were submitted to a nonparametric repeated measures ANOVA. The results were confirmed with a parametric repeated measures ANOVA. Significant differences (p 0.001) among the four groups with respect to dye penetration were detected, with the association Vitremer/Tetric showing the best results. The use of Vitremer in the open sandwich technique presents the lowest degree of microleakage among the treatments considered in this study.