Resin composite restorations in deep subgingival cavities involving cementum margins often exhibit early adhesive debonding and marginal failure. Conventional resin adhesive systems are prone to hydrolysis in these situations. However, evidence on the optimal adhesive systems for deep margins is limited. Therefore, this study evaluated the effect of HEMA-containing and HEMA-free adhesives in deep proximal margin restorations after simulated long-term aging.Class II box-only cavities were prepared on the distal surfaces of 48 extracted upper premolars, with the gingival margins placed 2 mm apical to the cemento-enamel junction. The teeth were randomly divided into three study groups (n = 16) based on the type of adhesive system: two-step HEMA-containing, two-step HEMA-free, and one-step HEMA-containing. Cavities were restored using a preheated nanohybrid resin composite and cured using a polywave LED light at 1200 mW/cm2. Each group was subdivided into an immediate subgroup (n = 8), in which teeth were examined after 24 hours, and a delayed subgroup (n = 8), in which teeth were examined after an aging process comprising 10,000 thermal cycles (5-55°C). Resin composite/cementum interfaces were evaluated quantitatively by scanning electron microscopy at 200× magnification. Marginal integrity was statistically analyzed by the percentage of continuous margin (%CM).The Kruskal-Wallis test revealed that the two-step HEMA-containing and two-step HEMA-free adhesive systems showed significantly better marginal integrity than the one-step HEMA-containing system for both the immediate and aged groups (all p < 0.005). A Wilcoxon rank-sum test showed that the aging process affected the one-step HEMA-containing adhesive system (p = 0.002), but not the two-step systems (p < 0.05).Two-step adhesive systems had better performance at cementum margins than the one-step adhesive system. The presence of HEMA in the two-step adhesive did not degrade marginal quality compared to the two-step HEMA-free adhesive systems.