Summary: The log-linear Birnbaum-Saunders model with $p$ independent variables is used for the study of fatigue in materials. A special variant of this model is obtained when the values of the independent variables are coded using a factorial design. In this case, the independent variables are orthogonal, allowing for more accurate parameter estimates. Another advantage of this model is that it is derived by considering the basic characteristics of the fatigue process. Therefore, it is more accurate for the study of fatigue in wood. We propose a multiple orthogonal polynomial Birnbaum-Saunders model to estimate fatigue in wood, with its respective maximum likelihood estimates and statistical inferences based on the asymptotic normal approximation of maximum likelihood estimates. Experimental data are presented on fatigue in wood under tension, obtained in the Laboratory of Wood and Timber Structures ‒ LaMEM-EESC-USP-Brazil. The estimation procedure based on the multiple orthogonal polynomial Birnbaum-Saunders model is compared with that obtained by considering the normal multiple orthogonal polynomial model. The results demonstrate that the parameter estimates obtained through the multiple orthogonal polynomial Birnbaum-Saunders model are at least 9\% more accurate. Therefore, this model can be used for the study of fatigue data in wood and its derivatives.