Prospective studies that have examined the association between physical activity and fracture risks have reported conflicting findings. We performed a meta-analysis to evaluate this association. We searched MEDLINE (1966 to February 1, 2013), EMBASE (1980 to February 1, 2013), and OVID (1950 to February 1, 2013) for prospective cohort studies with no restrictions. Categorical, heterogeneity, publication bias, and subgroup analyses were performed. There were 22 cohort studies with 1,235,768 participants and 14,843 fractures, including 8874 hip, 690 wrist, and 927 vertebral fractures. The pooled relative risk (RR) of total fractures for the highest versus lowest category of physical activity was 0.71 (95% confidence interval [CI], 0.63–0.80). The analysis of fracture subtypes showed a statistically significant inverse relationship between a higher category of physical activity and risk of hip and wrist fracture. The risk of hip or wrist fracture was 39% and 28% lower, respectively, among individuals with the highest category of physical activity than among those with the lowest category (95% CI, 0.54–0.69 and 0.49–0.96, respectively). The association between physical activity and vertebral fracture risk was not statistically related (RR, 0.87; 95% CI, 0.72–1.03). There was no evidence of publication bias. There was a statistically significant inverse association between physical activity and total fracture risk, especially for hip and wrist fractures. Additional subject-level meta-analyses are required for a more reliable assessment of subgroups and types of physical activity. © 2014 American Society for Bone and Mineral Research.