Purpose: This randomized, split-mouth clinical study evaluated the marginal quality of direct class-I and class-II restorations made of microhybrid composite that were applied using two polymerization protocols and two marginal evaluation criteria. Materials and Methods: A total of 50 patients (mean age: 33 years) received 100 direct class-I or class-II restorations in premolars or molars. Three calibrated operators made the restorations. After conditioning the tooth with 2-step etch-and-rinse adhesive, restorations were made incrementally using microhybrid composite. Each layer was polymerized using a polymerization device operated either in regular mode (600-650 mW/cm2 for 20 s) (RM) or high-power (1200-1300 mW/cm2 for 10 s) mode (HPM). Two independent, calibrated operators evaluated the restorations 1 week (baseline) and 6 months after restoration placement, and thereafter annually up to 10 years using modified USPHS and SQUACE criteria. Data were analyzed using the Mann-Whitney U-test (α = 0.05). Results: Alpha scores (USPHS) for marginal adaptation (76% and 74% for RM and HPM, respectively) and marginal discoloration (70% and 72%, for RM and HPM, respectively) did not show significant differences between the two polymerization protocols (p > 0.05). Alpha scores (SQUACE) for marginal adaptation (78% and 74% for RM and HPM, respectively) and marginal discoloration (70% for both RM and HPM) were also not significantly different at the 10-year year follow-up (p > 0.05). Conclusion: Regular and high-power polymerization protocols had no influence on the stability of marginal quality of the microhybrid composite tested up to 10 years. Both modified USPHS and SQUACE criteria confirmed that regardless of the polymerization mode, marginal quality of the restorations deteriorated significantly compared to baseline (p 0.05). Palabras clave: photopolymerization, resin composite, USPHS, SQUACE