A new medical device can take years to develop from early concept to product launch. The long development process can be attributed to the severe consequences for the patient if the device malfunctions. Three approaches are often combined to mitigate risks: rigorous simulation and modeling, physical test programs, and Failure Mode Effect Analysis (FMEA) — all of which are time-consuming. Physical test programs are often carried out on prototype components from the same batch and, therefore, limited in revealing the actual distribution of performance. The risk probabilities are subsequently based on educated guesses. Furthermore, simulation and modeling are usually performed on nominal geometry — not accounting for variation — and only provide a safety factor against failure. The traditional use of safety factors in structural analysis versus the probabilistic approach to risk management presents an obvious misfit. Therefore, these three approaches are not ideal for addressing the two key questions that the design engineer has: 1) How often will the design fail, and 2) How should the design be changed to improve robustness and failure rates. The present work builds upon the existing Robust and Reliability-Based Design Optimization (R2BDO) and adjusts it to address the key questions above using finite element analysis. The key feature of the new framework is the focus on minimal use of computational resources while being able to screen feasible design concepts early in the embodiment phase and subsequently optimize their probabilistic performance. A case study in collaboration with a medical design and manufacturing company demonstrates the new framework. The case study includes FEA contact modeling between two plastic molded components with twelve geometrical variables. The optimization focuses on minimizing the failure rate (and improving design robustness) concerning three constraint functions (contact pressure, strain, torque). The study finds that the new framework achieves significant improvements to the component’s performance function (failure rate) with minimal computational resources.