Some studies have indicated that the risk of fragility fractures in men increases as bone mineral levels decrease, but there is an overlap in the bone mineral density (BMD) measurements between patients with or without fractures. Furthermore, it has been suggested that the biomechanical competence of trabecular bone is dependent not only on the absolute amount of bone present but also on the trabecular microarchitecture. In the present study, 108 men (mean age 52.1 years) with lumbar osteopenia (T score < -2.5) were recruited to examine the relationships between BMD, architectural changes in trabecular bone, and the presence of vertebral fractures. Lumbar BMD was assessed from L2 to L4 in the anteroposterior view with dual-energy X-ray absorptiometry. At the upper left femur, hip BMD was measured at the transcervical site. Spinal X-ray films were analyzed independently by two trained investigators, and vertebral fracture was defined as a reduction of at least 20% in the anterior, middle, or posterior vertebral height. Transiliac bone biopsy specimens were obtained for all patients. Histomorphometric studies were performed on an image analyzer, and the following parameters were determined: trabecular bone volume (BV/TV), trabecular thickness (Tb.Th), number (Tb.N), and separation (Tb.Sp), interconnectivity index (ICI), characterization of the trabecular network (node count and strut analysis), and star volume of the marrow spaces. Spinal radiographs evidenced at least one vertebral crush fracture in 62 patients (group II) and none in 46 patients (group I). After adjusting for age, body mass index, and BMD, there were no significant differences between the two groups in BV/TV, Tb.Th, or star volume. In contrast, the mean values of ICI, free end-to-free end struts (FF/TSL), and Tb.Sp were significantly higher, whereas Tb.N and node-to-node struts (NN/TSL) were lower in patients with at least one vertebral fracture. Logistic regression analysis showed that only ICI, FF/TSL, NN/TSL, and Tb.N were significant predictors of the presence of vertebral fracture: odds ratios for an alteration of 1 SD ranged from 1.7 (1.0-3.2) for NN/TSL to 3.2 (1.1-10.1) for ICI. Patients with at least three vertebral fractures (n = 23) were categorized as "multiple fractures." The results of logistic regression showed that spine BMD, BV/TV, and all architectural parameters were significant predictors of multiple vertebral fractures: odds ratios for an alteration of 1 SD ranged from 2.2 (1.1-4.6) for star volume to 3.7 (1.4-9.7) for ICI. These results strongly suggest that bone trabecular microarchitecture is a major and independent determinant of vertebral fractures in middle-aged men with osteopenia.