The treatment for pelvic sarcoma often involves complete resection of the tumor and results in a sizable bone defect in the pelvis. Recently, reconstruction of the resulting bone defect with a 3D-printed custom-made implant has become increasingly popular because of the promising functional outcomes it can provide. However, the typical design process of a custom implant today lacks an engineering assessment of the implant’s durability. As a result, complications due to structural failures, such as fixation failures, remain high. Breakage or pullout failures of fixation screws are common, sometimes necessitating surgical revisions. With adequate design verifications, fixation failures may be adverted.

A few studies have incorporated patient-specific finite element models into the implant design process, in an attempt to detect potential structural failures of the postoperative pelvic construct. However, these finite element models are usually subject to arbitrary modeling choices, casting doubts about the reliability of these models. To date, no standard for constructing these finite element models has been proposed and the effect of many modeling choices on the predicted fixation durability remains unexplored.

This thesis presents a series of investigations of the effect of various boundary and loading conditions on computational fixation durability evaluation using patient-specific finite element models. The investigations aimed to 1) examine the accuracy of the common modeling practices, 2) propose modeling techniques that better reflect the physiological relationships within the postoperative pelvis, and 3) improve the current framework of constructing the finite element model for fixation durability evaluation. One by one, four distinct aspects of the boundary and loading conditions were studied by addressing the unique modeling challenges posed by two different implant designs – 1) the interactions between the remaining bone and implant, 2) the variety of the hip joint contact forces, 3) the modeling techniques of two distinct categories of orthopedic fixation screws, and 4) the inclusion of muscle forces. First, we established that the computational fixation durability was sensitive to the imposed boundary and loading conditions. The predicted likelihood for screw failure was more conservative when boundary conditions reflected early-stage osseointegration or loading conditions reflecting a wide range of daily activities were applied. Second, we explored the importance of choosing a physiological screw model for predicting screw failures and proposed a novel method of modeling compressive screws. The proposed method made assessing pullout failures for compressive screws possible. Lastly, we implemented all the previous improvements of the model and incorporated muscle forces into the finite element model. This patient-specific finite element model was subject to not only physiological boundary conditions but also postoperative muscle and hip joint contact forces which were predicted with a personalized neuromusculoskeletal model. We found the inclusion of muscle forces had a greater influence on pullout failure evaluation than breakage failure evaluation.

Through these investigations, this thesis demonstrated the importance of carefully selecting physiological boundary and loading conditions for analyzing the durability of the screws used to fixate custom pelvic implants. The analyses of the modeling practices introduced in the improved model were crucial steps for gaining confidence in the computational evaluation framework within the pelvic implant design community and for providing a higher standard of care for pelvic sarcoma patients.