ObjectiveTo compare mechanical performance and enamel-crack propensity of direct, semi-direct, and CAD/CAM approaches for large MOD composite-resin restorations. Methods45 extracted maxillary molars underwent standardized slot-type preparation (5-mm depth and bucco-palatal width) including immediate dentin sealing (Optibond FL) for the inlays (30 teeth). Short-fiber reinforced composite-resin (EverX Posterior covered by Gradia Direct Posterior) was used for the direct approach, Gradia Direct Posterior for the semi-direct, and Cerasmart composite resin blocks for CAD/CAM inlays. All inlays were adhesively luted with light-curing composite-resin (preheated Gradia Direct Posterior). Shrinkage-induced enamel cracks were tracked by transillumination photography. Cyclic axial isometric chewing (5-Hz) was simulated, starting with a load of 200N (5000 cycles), followed by stages of 400, 600, 800, 1000, 1200, and 1400N (maximum 30,000 cycles each) until fracture or to a maximum of 185,000 cycles. Survived specimens were subjected to cyclic-load-to-failure test at 30-degree angle on the palatal cusp. ResultsOnly small shrinkage-induced cracks were found in 47% of the direct restorations compared to 7% and 13% of semi-direct and CAD/CAM inlays, respectively. Survival to accelerated fatigue was similar for all three groups (Kaplan–Meier p>.05) and ranged between 87% (direct) and 93% (semi-direct and CAD/CAM). Cyclic-load-to-failure tests did not yield significant differences either (Life Table analysis, p>.05) with median values of 1675N for CAD/CAM inlays, 1775N for fiber-reinforced direct restorations and 1900N for semi-direct inlays. SignificanceAll three restorative techniques yielded excellent mechanical performance above physiological masticatory loads. Direct restorations performed as good as inlays when a short-fiber reinforced composite-resin base was used.