On the premise that bone response to exercise is locally controlled, we conducted a randomized trial to evaluate the effects of a 1-year training of psoas muscles (treatment group: TG) versus a 1-year training of deltoid muscles (control group: CG) on the lumbar trabecular bone mineral density (TBMD). TBMD was measured with computed tomography scan. Seventy-eight subjects were included and 67 completed the study. Intention to treat analysis revealed no significant change in TBMD from 0 to 12 months. Data analysis in the 67 remaining women, including both assiduous and nonassiduous subjects, revealed greater bone loss in CG than in TG although the difference was not significant. Similar analysis in a subgroup of subjects who performed the exercises assiduously (TG: n = 23, CG: n = 26) showed that the mean bone loss of all four vertebrae from 0 to 12 months was significantly greater in the CG (-8.87 +/- 12.75 mg/cm3, mean +/- SD) than in the TG (0.14 +/- 11.21 mg/cm3, mean +/- SD, P = 0.01). These results suggest that continuous 1-year psoas training can prevent lumbar bone loss in postmenopausal women and support the hypothesis of local action of physical activity.