A design matrix in response surface method (RSM) that satisfies the orthogonality is very useful because the mean square error can be minimized, so that the response surface is more precise. But the orthogonality of a second-order design matrix in conventional RSM cannot be satisfied. In this paper, a second-order orthogonal experimental design (SOED)-based RSM is proposed by considering the orthogonality of high-order design matrix. The SOED is constructed by changing the length of star points, and the main characteristic of SOED is that the design matrix is diagonal. When the high-order terms are considered in the SOED-based RSM, a globe optimal solution can be found. As the regression equation is determined, the reliability index can be analyzed by the normalized distance between the mean value of the performance function and the critical limit state of the safety factor. A practical large-scale landslide with two slip surfaces is taken to verify the applicability and precision of the proposed method in detail. It is found that the SOED-based response surface is more rigorous than the conventional RSM.