The purpose of this study was to develop a simple and easy predictive model of leg, spine and whole body bone mineral density (BMD) from anthropometric, physical fitness, body composition and quantitative ultrasound (QUS) variables. Participants were 138 Japanese overweight and obese men (50.9 ± 9.6 yr, body mass index [BMI] 29.1 ± 2.6 kg/m²). We measured anthropometric variables (height, weight, BMI, chest, waist, hip, upper arm, thigh circumferences), physical fitness (grip strength, side steps, vertical jump, forced vital capacity), body composition (fat-free mass) and QUS. BMD was measured by dual energy X-ray absorptiometry. Multiple linear regression analyses showed that all predictive models for BMD were significant. As a result, the predictive model for leg BMD showed the highest model fitting. The Bland & Altman approaches demonstrated the (positive or negative) systematic error even though most plots were placed within ideal range. Predictive model from physical fitness, body composition and QUS would be useful for estimating whole body and regional BMD. Because these predictive models are likely to have some systemic errors, further research is needed to improve the predictive accuracy.