Introduction: Postural instability during walking and tripping over obstacles are the main causes of falls in people with Parkinson's disease (PD). Preliminary limited evidence suggests that the length of the prospective follow-up period affects falls prediction in PD, with shorter periods leading to more accurate prediction. Thus, the primary aim of the present study was to test the performance of center of pressure (CoP) variables during obstacle crossing to predict fall risk in people with PD during subsequent periods of four, six, and 12 months. We also compared CoP variables during obstacle crossing between fallers and non-fallers. Methods: Forty-two individuals with PD, in mild to moderate stages, completed the baseline obstacle crossing assessment and reported falls for 12 months. Participants walked at their self-selected pace and were instructed to cross an obstacle (half knee height) positioned in the middle of an 8-m long pathway. A force platform was used to analyze CoP parameters of the stance phase of the trailing limb (most affected limb). The ability of each outcome measure to predict fall risk at four, six, and 12 months was assessed using receiver operating characteristic curve analyses. Results: Ten individuals (23.8%) were considered fallers at four months, twelve individuals (28.5%) at six months, and twenty-one individuals (50%) at 12 months. CoP amplitude and CoP velocity in the mediolateral direction significantly predicted fall risk at four, six, and 12 months. As judged by the area under the curve, mediolateral CoP velocity showed the best performance at four months, while mediolateral CoP amplitude showed the best performance at six months. Fallers presented greater values of mediolateral CoP velocity and amplitude than non-fallers. Conclusion: These findings suggest that mediolateral CoP velocity and amplitude during obstacle crossing might be useful to predict fall risk in people with PD. Therefore, larger studies are encouraged.