Statement of problem: Patients with partial tooth loss treated with implant-supported fixed partial dentures (FPDs) have difficulty using conventional mandibular advancement devices (MADs) because of the risk of side effects. Also, which design factors affect biomechanical stability when designing MADs with better stability is unclear. Purpose: The purpose of this finite element (FE) analysis study was to analyze the effect of the MAD design on biomechanical behavior and to propose a new design process for improving the stability of MADs. Material and methods: Each 3D model consisted of the maxillofacial bones, teeth, and implant-supported FPDs located in the left tooth loss area from the first premolar to the second molar and a MAD. Three types of custom-made MADs were considered: a complete-coverage MAD covering natural tooth-like conventional MADs, a shortened MAD excluding the coverage on the implant-supported FPD, and a newly designed MAD without anterior coverage. For the new MAD design, topology optimization was conducted to reduce the stress exerted on the teeth and to improve retention of the MAD. The new MAD design was finished by excluding the coverage of the maxillary and mandibular central incisors based on the results of the topology optimization. A mandibular posterior restorative force for a protrusion amount of 40% was used as the loading condition. The principal stress and pressure of the cancellous bone and periodontal ligaments (PDLs) were identified. Results: Considering the load concentration induced by the complete-coverage MAD, bone resorption risk and root resorption risk were observed at both ends of the mandibular teeth. The shortened MAD resulted in the highest stress concentration and pressure with the worst stability. However, in the case of the complete-coverage MAD, the pressure in the PDLs was reduced to the normal range, and the risk of root resorption was reduced. Conclusions: For patients with implant-supported FPDs, MAD designs with different extents of coverage had an influence on biomechanical behavior in terms of stress distribution in cancellous bone and PDLs. A MAD design without anterior coverage provided improved stability compared with complete-coverage or shortened designs. The presented method for MAD design, which combined FE analysis and topology optimization, could be effectively applied in the design of such improved MADs.