BMD measured with DXA is widely used in clinical practice to assess fracture risk and guide management. DXA can also assess hip geometry, including femoral neck width (FNW) and hip axis length (HAL), which have both been associated with increased risk for hip fracture independently from BMD. Our objective was to assess if FNW predicts hip fracture independently from other factors including HAL. We performed a retrospective cohort study using the Province of Manitoba BMD registry. The study population comprised 75 095 individuals (90.8% women), mean age 64.7 yr, with baseline hip BMD and hip geometry parameters. Linked health records were used to ascertain subsequent hospitalization with hip fracture as a primary diagnosis. During a mean follow-up of 8.3 (SD 5.1) yr, 2341 incident hip fractures were recorded. Each SD increase in age- and sex-adjusted FNW was associated with incident hip fracture (HR 1.15, 95% CI 1.10-1.19), which was unchanged after adjustment for height, weight, FN BMD, and clinical risk factors. However, FNW showed a significant positive correlation with HAL (r = 0.68). When further adjusted for HAL, FNW was no longer associated with increased risk for hip fracture (HR 0.98, 95% CI 0.94-1.03). A similar pattern was seen for FN, and intertrochanteric and non-hip fractures. In contrast, increased risk of hip fracture was consistently seen with each SD increase in HAL even after adjustment for all covariates including FNW (HR 1.35, 95% CI 1.28-1.42). In conclusion, FNW is a risk factor for hip fracture before but not after adjustment for HAL. HAL, on the other hand, robustly and independently predicts hip fracture, including both FN and trochanteric fractures.