This cross-sectional study covered 301 individuals over 70 years of age--207 women (W) and 94 men (M)--living in the city of São Paulo, Brazil. Our aims were to evaluate the prevalence of low bone mineral density (BMD) in this population and the possible factors that influence BMD. The subjects were submitted to a bone densitometry scan (DXA) to evaluate the BMD at lumbar spine (LS), femoral neck (FN), trochanter (T), total femur (TF) and total body composition. At the time, the participants filled in a questionnaire about lifestyle habits, diet and medical history, as well as having blood samples taken to check hormone and biochemical levels. Anthropometric parameters were measured. Osteopenia and osteoporosis were defined in accordance with the criteria suggested by the World Health Organization. In the different sites studied, the prevalence of osteopenia and osteoporosis varied, in men ranging 33.3-57.4% and 6.4-16.1%, respectively, and in women ranging 36.6-56.5% and 22.2-33.2%, respectively. Weight was the variable that most strongly correlated with BMD at the proximal femur in both sexes (men, r =0.44-0.52; women, r =0.48-0.52) and with BMD at LS in women (r =0.44). Height was the parameter that best correlated with BMD at LS in men (r =0.34). In men follicle-stimulating hormone, growth hormone and glycemia correlated with BMD at T and TF, while plasma albumin only correlated with BMD at T. In women glycemia correlated with BMD at LS, and follicle-stimulating hormone correlated with BMD at FN, T and TF. In conclusion, we found a high prevalence of osteopenia and osteoporosis in this population, with weight being the best predictor of BMD. The prevalence of osteoporosis and osteopenia at FN was as high in men as that observed in women.