Purpose: To compare the effectiveness of different methods of fractured screw fragment retrieval from dental implants, comparing a conventional method (use of an exploration probe and ultrasonic tips) with two mechanical retrieval kits. Materials and Methods: One hundred thirty-five screws in implant internal connections were fractured to test fragment retrieval comparing three methods: conventional method with the use of an explorer probe and ultrasonic tips (group CE), extraction with a mechanical kit with a guide cylinder (group MK1), and extraction with a mechanical kit without a guide cylinder (group MK2). Extraction success and retrieval times for each method, as well as success and retrieval times in relation to fracture depth-coronal, middle, or apical-were analyzed by means of chi-square, Fisher exact, Kruskal-Wallis, and Mann-Whitney tests (P .05). The integrity of implant internal threads was also analyzed. Results: The extraction success rates for groups MK2, CE, and MK1 were 93.3%, 73.3%, and 20%, respectively. The mean ± SD extraction times were 2 minutes, 56 seconds ± 2 minutes, 9 seconds; 3 minutes, 15 seconds ± 1 minute, 48 seconds; and 13 minutes, 34 seconds ± 7 minutes, 25 seconds for groups MK2, CE, and MK1, respectively. Statistically significant differences were found in the number of fractured screws extracted and in the retrieval time. No significant differences were found for retrieval time or extraction success in relation to the depth of the fracture. Conclusion: The mechanical kit without the guide cylinder achieved the best results with respect to the number of screw fragments retrieved, retrieval time, and preservation of implant internal threads. Keywords: fragment screw, implant, mechanical kits, prosthetic