Model-assisted probability of detection (MAPOD) and sensitivity analysis (SA) are widelyused for measuring the reliability of nondestructive testing (NDT) systems., such as ultrasonictesting (UT), and understanding the effects of uncertainty parameters. In this work, a stochastic expansion-based metamodel is used in lieu of the physics-based NDT simulation model for efficient uncertainty propagation while keeping satisfactory accuracy. The proposed stochasticmetamodeling approach is demonstrated for MAPOD and SA on a benchmark case for UT simulations on a fused quartz block with a spherically-void defect. The proposed approach is compared with direct Monte Carlo sampling (MCS), and MCS with Kriging metamodels. The results indicate that around one order of magnitude reduction in the number of model evaluations required for MAPOD analysis can be obtained. Moreover, the results indicate around two orders of magnitude reduction of the number of model evaluations for the convergence of the statistical moments and obtaining the problem sensitivities.