Owing to the randomness of shrinkage and creep of concrete, random analysis method should be adopted to give a reasonable result with probabilistic guarantee. Stochastic variables for long-term analysis with GL2000 model about creep and shrinkage of concrete were presented. Combining Monte Carlo sampling with the response surface method, the stochastic analysis model was performed which is on time-variant deflection of prestressed concrete bridge. The long-term deflection of a prestressed concrete bridge was analyzed by this model. Sensitivity analysis and parametric study were carried out. And the uncertainty of creep model, elastic modulus of concrete, magnitude of sustained load, and prestress forces are the most important factors for long-term deflection. Furthermore, controlling accuracy of prestress forces and delaying the loading time are helpful to inhibit long-term deflection.