The upper body accelerations of people with Parkinson's disease (PD) measured by inertial measurement units (IMUs) may contribute towards diagnostic algorithms and help track disease progression. Before extracting variables related to upper body motion, acceleration signals require realignment to a global reference; however, the impact of these techniques on the resulting upper body variables is unclear. Therefore, the aim of this investigation was to examine the impact of four different realignment methods designed to correct acceleration signals on a range of upper body variables in older adults and in patients with PD. Two minutes of continuous gait were measured in 54 community-dwelling older adults (71.1 ± 6.7 years) and 60 people with PD (age: 68.5 ± 9.1 years). Three IMUs placed on the 5th lumbar vertebra, 7th cervical vertebra and the back of the head recorded the acceleration of the upper body. A selection of upper body variables sensitive to impaired upper body control in PD and four acceleration realignment methods were compared. A mixed-model ANOVA showed that the choice of realignment method significantly affected the values of upper body variables as well as their ability to discriminate between the PD and control group. Our findings indicate researchers and clinicians should be cautious when comparing upper body variables extracted from IMUs using different realignment methods, and consideration of realignment technique will be important when identifying the most sensitive markers of disease presence and progression. Therefore, it's strongly recommend that researchers consider and report their realignment methods when assessing upper body variables during gait.