BACKGROUND: In pre-operative planning for total hip arthroplasty (THA), femoral offset (FO) is frequently underestimated on AP pelvis radiographs as a result of inaccurate patient positioning, imprecise magnification, and radiographic beam divergence. The aim of the present study was to evaluate the accuracy and reliability of predicting three-dimensional (3-D) FO from standardised AP pelvis radiographs. METHODS: In a retrospective cohort study, pre-operative AP pelvis radiographs, AP hip radiographs and CT scans of a consecutive series of 345 patients (345 hips, 146 males, 199 females, mean age 60 (range: 40-79) years, mean body-mass-index 27 (range: 19-57)kg/m(2)) with primary end-stage hip OA were reviewed. Patients were positioned according to a standardised protocol and all images were calibrated. Using validated custom programmes, FO was measured on corresponding radiographs and CT scans. Measurement reliability was evaluated using intra-class-correlation-coefficients. To predict 3-D FO from AP pelvis measurements and to assess the accuracy compared to CT, the entire cohort was randomly split into subgroups A and B. Gender specific regression equations were derived from group A (245 patients) and the accuracy of prediction was evaluated in group B (100 patients) using Bland-Altman plots. RESULTS: In the entire cohort, mean FO was 39.2mm (95%CI: 38.5-40.0mm) on AP pelvis radiographs, 44.1mm (95%CI: 43.4-44.9mm) on AP hip radiographs and 44.6mm (95%CI: 44.0-45.2mm) on CT scans. In group B, we observed no significant difference between gender specific predicted FO (males: 48.0mm, 95%CI: 47.1-48.8mm; females: 42.0mm, 95%CI: 41.1-42.8mm) and FO as measured on CT (males: 47.7mm, 95%CI: 46.1-49.4mm, p=0.689; females: 41.6mm, 95%CI: 40.3-43.0mm, p=0.607). CONCLUSIONS: The present study suggests that FO can be accurately and reliably predicted from AP pelvis radiographs in patients with primary end-stage hip osteoarthritis. Our findings support the surgeon in pre-operative templating on AP-pelvis radiographs and may improve offset and limb length restoration in THA without the routine performance of additional radiographs or CT.