The goal was to compare the reliability of a novel 3D method with the standard 2D technique for lower limb discrepancy (LLD) measurement during pre-operative THA planning. This prospective study included 100 consecutive patients who underwent THA using 3D planning based on a low-dose CT scan. The LLD was subdivided into three parameters: the intra-articular LLD (IA-LLD), the segmental extra-articular LLD (EA-LLD), and the total LLD (T-LLD). The LLD was assessed with a standard 2D technique on CT scanograms and also with a 3D method. A pelvic reference line (PBL) was determined as the 3D line joining the deepest part of the two great sciatic notches. The IA length was measured from the lesser trochanters (MLT) to the PBL. The EA length was measured from the MLT to the ankle center, and the total length was measured from the ankle center to the PBL. The intra- and inter-observer reliability of the measurements was assessed with the intra-class correlation coefficient (ICC). The intra-observer ICC was higher with the 3D technique for IA-LLD (0.96-0.97 vs. 0.79-0.84), EA-LLD (0.96 vs. 0.78-0.92), and T-LLD (0.99 vs. 0.90-0.97). Inter-observer ICC was also higher with the 3D technique for IA-LLD (0.90-0.94 vs. 0.70-0.84) and EA-LLD (0.93-0.96 vs. 0.80-0.82), but not for T-LLD (0.91-0.94 vs. 0.91-0.94). The presented 3D method has a higher reliability than 2D assessment of LLD during pre-operative THA planning. This article presents the first discussion of measuring LLD from 3D models. As 3D reconstruction becomes both more feasible and less-invasive, this study has interest to the orthopaedic surgeon.