Modular augmented stems of a constrained condylar knee implant are intended to improve tibial fixation under increased varus/valgus loads, but conflicting studies have not yet indicated the factors determining stem usage and performance. To address this, we combined a paired-tibiae, cadaveric experiment of unstemmed and stemmed tibial components with specimen-specific computational models. We hypothesized that the stem would improve implant stability by decreasing implant motion and compressive strains in the proximal cancellous bone due to load transfer by the stem. The models also would indicate the important factors governing stem performance. Large variations of the displacements arose because of loading and biologic variability indicating the inconclusive effects of a stem. Despite these variations, the models showed that a stem augment consistently decreased the strains (30%-50%) in the bone beneath the tray. In tibiae of sufficient stiffness, the supporting cancellous bone did not approach yield, suggesting that a stem augment may not always be necessary. On the other hand, tibial specimens with reduced bone quality and lower stiffness benefited from a stem augment that transferred load to the distal cortical bone. Therefore, patient selection and proper sizing of the implant were identified as important factors in the analyses.