Estimation of hip joint loading is fundamental for understanding joint function, injury and disease. To predict patientspecific hip loading, a musculoskeletal model must be adapted to the patient’s unique geometry. By far the most common and cost effective clinical images are whole pelvis plain radiographs. This study compared the accuracy of anisotropic and isotropic scaling of musculoskeletal model to hip joint force prediction by taking patient-specific bone geometry from standard anteroposterior radiograms. Methods: 356 hips from 250 radiograms of adult human pelvis were analyzed. A musculoskeletal model was constructed from sequential images of the Visible Human Male. The common body position of one-legged stance was substituted for the midstance phase of walking. Three scaling methods were applied: a) anisotropic scaling by interhip separation, ilium height, ilium width, and lateral and inferior position of the greater trochanter, b) isotropic scaling by pelvic width and c) isotropic scaling by interhip separation. Hip joint force in one-legged stance was estimated by inverse static model. Results: Isotropic scaling affects all proportions equally, what results in small difference in hip joint reaction force among patients. Anisotropic hip scaling increases variation in hip joint force among patients considerably. The difference in hip joint force estimated by isotropic and anisotropic scaling may surpass patient’s body weight. Conclusions: Hip joint force estimated by isotropic scaling depends mostly on reference musculoskeletal geometry. Individual’s hip joint reaction force estimation could be improved by including additional bone geometrical parameters in the scaling method.