To evaluate the effect of ceramic thicknesses, bonding surface (enamel vs. dentin), and preparation design (box vs. no box) on the fatigue survival and failure load of minimally invasive full-veneer restorations. Human-premolars (n=60) were divided into five test groups (n=12). All teeth received full-veneer preparation with the following occlusal/labial thicknesses: standard: 1.5/0.8 mm; thin: 1.0/0.6 mm; ultrathin 0.5/0.4 mm. Preparations for each ceramic thickness were refined in enamel (E-1.0 and E-0.5) or dentin (D-1.5, D-1.0, and D-0.5). Control groups DB-1.5, EB-1.0, and EB-0.5 received box preparations. Monolithic lithium disilicate restorations (IPS-e.max-Press, Ivoclar Vivadent) were adhesively cemented (Syntac-Classic/Variolink-II, Ivoclar Vivadent) and subjected to cyclic mechanical loading (F=49 N, 1.2 million cycles) with simultaneous thermocycling (5-55°C). All specimens were exposed to single load-to-failure. Pair-wise differences were calculated by using a linear regression model and Student-Newman-Keuls method (p < 0.05). All full-veneers of group D-1.5, E-1.0, E-0.5, DB-1.5, EB-1.0, and EB-0.5 survived fatigue. Two full-veneers (D-1.0 and D-0.5) revealed cracks during fatigue, resulting in an overall fatigue survival rate of 98.1%. Mean load-to-failure values (N) were as followed: 1005 (D-1.5); 866 (D-1.0); 816 (D-0.5); 1495 (E-1.0); 1279 (E-0.5); 1129 (DB-1.5); 1087 (EB-1.0); and 833 (EB-0.5). Irrespective of ceramic thicknesses, enamel-based full-veneers resulted in higher failure loads than dentin-based restorations. Box preparation reduced the failure loads of thin and ultrathin enamel-based restorations. All tested monolithic lithium disilicate full-veneer restorations exceeded physiological masticatory forces. Minimally invasive full-veneer restorations with enamel as a bonding surface and a non-retentive preparation design showed superior performance. Enamel-based non-retentive full-veneers made of monolithic lithium disilicate may serve as a reliable and esthetical minimally invasive treatment option for premolars.