The scope of this paper is to assess the performance and computational effort of stochastic simulation methods for evaluating the probability of failure of bridges subjected to high-speed trains. Bridge and train represent an uncertain interacting dynamic system, and application of crude Monte Carlo simulation to a realistic mechanical model of this system is prohibitively expensive. Applicability and efficiency of two alternative stochastic methods, i.e. line sampling and subset simulation, to this problem is evaluated. The studies are conducted on a simplified mechanical model, composed of a plain beam representing the bridge and a planar mass-springdamper system representing the train. Such a computationally efficient model captures the fundamental characteristics of dynamic bridge-train interaction, and thus, facilitates the assessment of the stochastic methods. The results show that at a single train speed both line sampling and subset simulation reduce significantly the computational effort for predicting the probability of failure of bridges, whose random response is dominated by resonance effects.