Objective To investigate whether body mass index (BMI) affects the accuracy of vertebral bone quality (VBQ) in determining bone mineral density (BMD) in patients undergoing lumbar fusion surgery. Methods In this retrospective study, patients with preoperative noncontrast T1-weighted MRI were included. Restricted Cubic Spline (RCS) was employed to explore the nonlinear relationship between BMI and VBQ. Then patients were stratified according to the threshold of BMI. Pearson correlation analysis and linear regression were used to analyze the correlation between VBQ and the BMD in different groups. Receiver operating characteristic (ROC) analysis to calculate the area under the curve (AUC) was used to assess diagnostic efficacy according to BMI. Results A total of 328 patients (201 female and 127 male patients) with a mean age of 68.3 +/- 3.3 years were included in the study. Significant nonlinear relationship was observed given the results of RCS. In patients with BMI < 23.8 kg/m2, the correlation coefficient between VBQ and the lowest BMD was - 0.32 and significant distribution difference of VBQ score was observed between osteoporosis and normal as well as osteopenia subgroups. However, in patients with BMI >= 23.8 kg/m2, the correlation coefficient between VBQ and the lowest BMD was - 0.39 and significant distribution difference of VBQ score was observed in all three subgroups. In addition, the ROC analysis revealed that the predictive performance in determining low BMD was superior in patients with BMI >= 23.8 kg/m2 (AUC 0.80 vs. AUC 0.66, p = 0.034). Conclusions In this study, significant nonlinear relationship between BMI and VBQ was observed. Compared with patients with BMI < 23.8 kg/m2, VBQ has better discrimination between higher BMI (>= 23.8 kg/m2) patients with low BMD and those with normal bone density.