Aiming at the sensitivity analysis of probability box with both aleatory and epistemic uncertainties, a global sensitivity analysis method using the overlap area before-and-after probability box pinching as an uncertainty measure is proposed. Aleatory and epistemic uncertainties are widely existed in aerospace systems and the probability box method has been widely used in quantifying these uncertainties. The theory of an uncertainty reduction method is introduced for traditional probability box sensitivity analysis. Based on this theory, the position and shape of the offset of probability box is further considered. The influence degree of each input uncertainty is characterized by calculating the overlap of probability box area before-and-after reduction. The detailed implementation steps are described. Through numerical examples, the global sensitivity of the proposed method is compared with that of the traditional method. Finally, the proposed method is applied to the sensitivity ranking of an aero engine overall performance simulation. The results show that the proposed method is more applicable than the traditional pinching method, and the results are more accurate.