Objectives To examine the extent to which geometric parameters derived from dual-energy x-ray absorptiometry (DXA) scans in the UK Biobank (UKB) study are related to hip osteoarthritis (HOA) independently of sex, age and body size. Methods Femoral neck width (FNW), diameter of the femoral head (DFH) and hip axis length (HAL) were derived automatically from left hip DXA scans in UKB using outline points placed around the hip by a machine-learning program. Correlations were calculated between geometric parameters, age, height, and weight. Logistic regression was used to examine the relationship of geometric parameters with radiographic hip osteoarthritis (rHOA) and hospital diagnosed HOA (HESOA), and Cox proportional hazards model to evaluate the relationship with total hip replacement (THR). Analyses were adjusted for sex, age, height, weight, and geometric parameters. Results Complete data were available for 40,312 participants. In age and sex-adjusted analyses, FNW, HAL and DFH were all related to increased risk of rHOA. Despite strong relationships between geometric parameters and body size, relationships between geometric parameters and HOA showed little attenuation after adjustment for height and weight. Following mutual adjustment, both HAL and FNW retained independent relationships with rHOA, while DFH was now protective. Only FNW was independently related to HESOA and THR. Conclusion Greater FNW and HAL were independently related to an increased risk of rHOA, whereas greater DFH appeared to be protective. Greater FNW was independently predictive of HESOA and THR. These results suggest DXA-derived geometric parameters, particularly FNW, could help to predict HOA and THR risk.