For failure probability analysis of complex structures, the Monte Carlo method (MC) needs alarge number of samples and it is more inefficient in the case of small failure probability. Basedon the Subset Simulation (SS) and Subset Simulation Significant Sampling (SSIS), this paperanalyzes the static and dynamic reliability of main girder of cable-stayed bridge and realizesthe efficient solution of the failure probability of complex cable-stayed bridge. According toSS, MC sampling is performed on random variables. Finally, the critical values of intermediatefailure events are solved and automatically stratified, and then the failure probability of eachfailure mode is obtained by counting the number of samples in the failure domain. Accordingto SSIS method, SS method is optimized by introducing important sampling density function,which can make the sample points to fall in the structural failure areas as many as possible. The failure probability of main girder of cable-stayed bridge can be calculated more efficiently. Meanwhile, ANSYS-PDS template technique is used to perform the reliability and sensitivityanalysis of main girder of cable-stayed bridge. The failure probabilities obtained are very closeto the results of SS. The example analysis shows that for the reliability analysis of cable-stayedbridges, the sample size of SS and SSIS is 0.3% of MC, and the maximum relative difference isless than 3%. In addition the main beam section bending moment and wind speed have thegreatest impact on the reliability of the main beam of cable-stayed bridges.