The goal of this study was twofold. First, we aimed to evaluate the accuracy of a finite element (FE) model to predict bone fracture in cancer patients with proximal femoral bone metastases. Second, we evaluated whether femoroplasty could effectively reduce fracture risk. A total of 89 patients were included, with 101 proximal femurs affected with bone metastases. The accuracy of the model to predict fracture was evaluated by comparing the FE failure load, normalized for body weight, against the actual occurrence of fracture during a 6-month follow-up. Using a critical threshold, the model could identify whether femurs underwent fracture with a sensitivity of 92% and a specificity of 66%. A virtual treatment with femoroplasty was simulated in a subset of 34 out of the 101 femurs;​ only femurs with one or more well-defined lytic lesions were considered eligible for femoroplasty. We modeled their lesions, as well as the surrounding 4 mm of trabecular bone, to be augmented with bone cement. The simulation of femoroplasty increased the median failure load of the FE model by 57% for lesions located in the head/neck of the femur. At this lesion location, all high risk femurs that had fractured during follow-up effectively moved from a failure load below the critical threshold to a value above. For lesions located in the trochanteric region, no definite improvement in failure load was found. Although additional validation studies are required, our results suggest that femoroplasty can effectively reduce fracture risk for several osteolytic lesions in the femoral head/neck.