Background The unipodal stance task is a clinical task that quantifies postural stability and alignment of the lower limb joints, while weight bearing on one leg. As persons with knee osteoarthritis (KOA) have poor postural and knee joint stability, objective assessment of this task might be useful. Objective To investigate the discriminant validity of three-dimensional joint kinematics and centre of mass displacement (COM) between healthy controls and persons with knee KOA, during unipodal stance using inertial sensors. Additionally, the reliability, agreement and construct validity are assessed to determine the reproducibility and accuracy of the discriminating parameters. Methods Twenty healthy controls and 19 persons with unilateral severe KOA were included. Five repetitions of the unipodal stance task were simultaneously recorded by an inertial sensor system and a camera-based system (gold standard). Statistical significant differences in kinematic waveforms between healthy controls and persons with severe knee KOA were determined using one-dimensional statistical parametric mapping (SPM1D). Results Persons with severe knee KOA had more lateral trunk lean towards the contralateral leg, more hip flexion throughout the performance of the unipodal stance task, more pelvic obliquity and COM displacement towards the contralateral side. However, for the latter two parameters the minimum detectable change was greater than the difference between healthy controls and persons with severe knee KOA. The construct validity was good (coefficient of multiple correlation 0.75, 0.83 respectively) and the root mean squared error (RMSE) was low (RMSE <1.5°) for the discriminant parameters. Conclusion Inertial sensor based movement analysis can discriminate between healthy controls and persons with severe knee KOA for lateral trunk lean and hip flexion, but unfortunately not for the knee angles. Further research is required to improve the reproducibility and accuracy of the inertial sensor measurements before they can be used to assess differences in tasks with a small range of motion.