Because the main complication of implant-supported prostheses is torque loosening and/or fixation screw fracture, the goal of this study was to evaluate the torque before and after fatigue (screw placement and removal, respectively), single load-to-failure (compression test), and stress concentration of straight and angled abutments. Eighty implants were included in polyurethane cylinders. Half of the implants received straight abutments (group S, n = 40) and the other half received angled abutments (group A, n = 40). The abutments for cemented prostheses were installed with a torque of 20 Ncm. Eighty titanium structures were machined and cemented on the abutments with zinc-phosphate cement. After storage for 24 hours, half of the specimens had their torque loosening evaluated and were then immediately submitted to a compressive test in a universal testing machine (1 mm/minute, 1,000 kgf), while the other half were subjected to cyclic fatigue (200 N at 2 Hz for 2 × 106 cycles at 37°C) as an aging protocol (n = 20 from each group). The aged samples then had their torque loosening measured and were also submitted to the compression test. Representative samples were evaluated by scanning electron microscopy. Two bidimensional models similar to the in vitro specimens were created and analyzed using the finite element method to evaluate the stress concentration. Data from the in vitro tests were submitted to two-way analysis of variance and Tukey test, both with significance at P = .5. The results show that angled abutments are less capable of maintaining the installation torque and are less resistant during the single load-to-failure test. The von Mises stress concentration was higher for group A in the cervical region. The straight abutments have better prognosis than angled abutments and less susceptibility to mechanical failures.