The relationship between iron and bone mineral density (BMD) is still poorly understood. We investigated the associations of iron intake, serum iron and serum ferritin with BMD. This cross-sectional study identified 4000 females aged 12 to 49 years with complete and valid data on iron intake, serum iron, serum ferritin, and femoral neck and lumbar spine BMD from the National Health and Nutrition Examination Survey 2005–2010. Daily iron intake was the mean intake of iron nutrient ascertained from two consecutive 24-h dietary recalls;​ serum iron and serum ferritin were directly measured with established methods. Femoral neck and lumbar spine BMD were measured by Dual-energy X-ray absorptiometry (DXA). After adjusting for multiple covariates (i.e., age, body mass index and race), we used linear regression and generalized additive models (GAMs) to test the linear and non-linear associations of iron intake, serum iron and serum ferritin with BMD. The mean age of this study was 27.70 years (SD = 11.88 years). Higher serum ferritin was associated with lower femoral neck and lumbar spine BMD (all adjusted P < 0.05);​ iron intake and serum iron were not associated with femoral neck and lumbar spine BMD. Similar results were found when iron levels were classified as iron deficiency, normal iron and iron overload. There were no obvious non-linear relationships between the above three iron variables and BMD in the GAM analyses. There was a negative and linear association between serum ferritin and BMD;​ iron intake and serum iron were not associated with BMD. Serum ferritin appeared to be a better iron variable than iron intake and serum iron in relation to BMD.