The bone quality has a significant effect on the insertion torque of dental implant. In most clinical studies, bone density is used as a gold standard in predicting insertion torque. By contrast, trabecular microstructure is ignored. In this study, a microstructure-based numerical computational method with high accuracy and efficiency for the insertion torque of dental implant was proposed by introducing two microscopic variables, namely, volume fraction and fabric tensor. First, two kinds of 3D microstructural solid models with same volume fraction and fabric tensor were established on the basis of the microstructural topology of six reference specimens. Second, a new numerical simulation method based on homogenous theory was used to explore the material models of these 3D microstructural solid models at the microscopic scale. Then, the anisotropic material models of specimens were developed on the basis of the mixture rule. Thereafter, a numerical simulation based on the anisotropic finite element (FE) model was carried out to acquire the insertion torque. To demonstrate the efficiency and accuracy of the simulation based on the anisotropic FE model, numerical simulations based on isotropic FE model and micro-computer tomography (micro-CT) FE models were also implemented as comparisons. Comparison of the simulated peak insertion torques of the anisotropic, isotropic, and micro-CT FE models with insertion experiments demonstrated the feasibility and potential of the proposed method. The anisotropic FE model reduced the time consumption by 91.85% and enhanced the accuracy by 11.82% compared with the micro-CT and isotropic FE models, respectively. Copyright © 2019 Elsevier Ltd. All rights reserved.