Objective: The aim of this study was to compare the marginal microleakage between bulk-fill, preheated bulk-fill, and bulk-fill flowable composite resins above and below cemento-enamel junction (CEJ) using micro-computed tomography. Methods: Sixty freshly extracted premolar teeth were prepared with a slot shaped cavities of a total of 120 Class II: 3mm (bucco-lingual), 2mm (mesio-distal) with mesial-gingival margin located 1mm coronal to CEJ, and distal gingival margin located 1mm apical to the CEJ. The samples were randomly divided into two main groups according to the restorative material (Tetric EvoCeram and 3M Filtek), and each group was further sub-divided into three subgroups according to the consistency (bulk fill, preheated bulk fill, and bulk fill flowable). The specimens were thermocycled (500 cycles,5–55 °C, 30 seconds dwell time) and immersed in a 50% ammonical silver nitrate for 24 hours. Microleakage was evaluated using micro-CT. Data analyzed via three-way ANOVA and random forest model (a machine learning algorithm). Results: There was no statistically significant three-way interaction between type of composite, location of gingival margin related to CEJ, and filling technique (P=0.103). However, the random forest model (a machine learning algorithm) showed type of composite is the most influential factor. Conclusion: Within the limitations of this study, considering the importance of the type of composite, the teeth restored with preheated bulk fill composite showed decrease in microleakage in comparison with bulk fill flowable composite
