Purpose: To compare the marginal adaptation of adhesively luted ceramic inserts in standardized cylindrical cavities of bovine dentin and enamel with the marginal adaptation of adhesively luted ceramic inlays in extracted molars after mechanical loading, and to calculate the minimum sample size required to differentiate between material groups based on the findings of this investigation. Materials and Methods: Ceramic inserts of similar dimension were luted in standardized cylindrical cavities of bovine dentin and enamel (n = 6, Ø = 4 mm) with seven different adhesive/resin cement systems (Syntac/Variolink, Prime&Bond NT/Variolink, Excite DSC/Variolink, AdheSE/Variolink, Excite DSC/Multilink, Multilink Primer/Multilink, RelyX Unicem). The same materials were used to lute ceramic inlays (Empress II) in three-surface cavities of extracted human molars (n = 6 per group). All specimens were submitted to 2000 cycles of thermocycling. In addition, restored teeth were submitted to cyclic loading (640,000 cycles, 50 N) in a chewing simulator. Replicas after stressing were analyzed with SEM, and the percentage of continuous margin of the inserts and the inlays was calculated, differentiating the proximal part of the inlay into cervical dentin/enamel and axio-proximal enamel. The mean percentage values per material group of the insert and inlay groups were ranked using relative ranks. Sample size estimation was done for pooled standard deviations comparing between two and seven materials and assuming 20% or 10% of the mean of continuous margin to be statistically significant. Results: The Spearman correlation coefficient between the variables "insert dentin" and "inlay cervical dentin" was 0.71 (p = 0.07), between "insert enamel" and "inlay axio-proximal enamel" 0.07 (p = 0.9). The variability of the test results was large for both the insert and inlay variables, especially at the resin-dentin interface. No statistically significant difference between the materials could be found for the insert method when ANOVA and Bonferroni post-hoc tests were applied (p > 0.05), while the inlays luted with Prime&Bond NT/Variolink showed significantly less continuous margin at the cervical dentin than all other groups with the exception of RelyX Unicem. At least 16 (63) inlays and 14 (57) inserts had to be used per group for differences of 20% (10%) of the mean of continuous margin. Conclusion: To evaluate luting agents with regard to their ability to reduce marginal discrepancies, the inlay model is inadequate due to high sample numbers which make the test time-consuming and expensive; furthermore, its clinical relevance is uncertain. Alternatively, the insert method may be a suitable screening method for dentinal margins, although its clinical relevance is also unknown. Keywords: marginal adaptation, SEM, luting cements, ceramic inlay, ceramic insert, sample size