Background: and Purpose Falls are not uncommon even in patients with early stages of Parkinson’s disease (PD). The aims of this study were to determine the relationships between gait parameters and falls and identify crucial gait parameters for predicting future falls in patients with de novo PD. Methods: We prospectively recruited patients with de novo PD, and evaluated their baseline demographics, global cognitive function on the Montreal Cognitive Assessment test, and parkinsonian motor symptoms including their subtypes. Both forward gait (FG) and backward gait (BG) were measured using the GAITRite system. The history of falls in consecutive patients with de novo PD was examined along with 1 year of follow-up data. Results: Among the 76 patients with de novo PD finally included in the study, 16 (21.1%) were classified as fallers. Fallers had slower gait and shorter stride for FG and BG parameters than did non-fallers, while stride-time variability was greater in fallers but only for BG. Multivariable logistic regression analysis revealed that slow gait was an independent risk factor in BG. Conclusions Among the patients with de novo PD, gait speed and stride length were more impaired for both FG and BG in fallers than in non-fallers. It was particularly notable that slow BG was significantly associated with future fall risk, indicating that BG speed is a potential biomarker for predicting future falls in patients with early-stage PD.