The objective of this study was to investigate the incidence of vertebral fractures (VFx) and the value of physical function (PF) and body composition (BC) for predicting VFx in a Japanese population. This study included 307 subjects (113 men, 194 women) at least 40years of age who were assessed at community health check-ups in 2008 and 2016. PF was assessed by grip strength and by single-leg stance, timed up-and-go, and 30-s chair stand tests, each scored from 0 to 3 for a possible total of 12 points (higher scores reflect lower function). BC was scored on bioelectrical impedance measurements of trunk and appendage muscle volume, with 6 possible points. We diagnosed radiological VFx semiquantitatively on lateral views of the lumbar spine, and measured bone mineral status by quantitative ultrasound (QUS) of the calcaneus. We conducted logistic regression analysis with VFx as the dependent variable and age, sex, BMI, QUS, PF score, and BC score as independent variables. In 8years, 36 participants (12%) sustained new VFx. After correcting for age, sex, BMI, and QUS, the odds of VFx increased with a PF score ≥ 8 (OR 5.6; 95% CI 1.21-25.90; P = 0.028) and increased further with a PF + BC score ≥ 9 (OR 8.1; 95% CI 1.80-36.00; P < 0.01). Both PF and BC are important for predicting fragility fractures. The scoring system used here may reflect small differences better than categorical (single cutoff) definitions of poor function.