Bone microarchitecture in osteoporosis can be characterized by examining iliac bone biopsies and treatment effects assessed by comparing a baseline biopsy from one side to a posttreatment biopsy from the other side, a method that assumes limited side-to-side variability. New techniques based on micro-computed tomography (μCT) provide information on the three-dimensional (3D) microarchitecture of bone. We used μCT to measure side-to-side and within-side variability of 3D microarchitectural parameters of trabecular and cortical bone in paired iliac-crest biopsies, one from each side. A Bordier needle trephine was used to collect biopsies from 30 postmenopausal female cadavers (mean age, 73.7 ± 10.7 years;​ range, 55–96 years). Biopsies were chemically defatted then imaged using a desktop μCT scanner (voxel size, 10.77 μm). Parameters measured in trabecular bone consisted of bone volume/tissue volume (BV/TV, %), direct trabecular thickness and trabecular spacing (Tb.Th* and Tb.Sp*, μm) using the sphere method, bone surface/bone volume (BS/BV, mm⁻¹), trabecular number (Tb.N, mm^-1), structure model index (SMI), trabecular pattern factor (Tb.Pf), and degree of anisotropy (DA). In cortical bone, we measured cortical thickness (Cort.Th), porosity (Cort.Porosity), and pore diameter (Po.Dm). For trabecular bone parameters, reproducibility as assessed from two μCT acquisitions ranged from 4.1% to 6.9%. To assess side-to-side variability, we matched the volumes of interest selected in the right and left iliac crests. The mean difference in absolute individual percent variation (mAbsΔ_ind) between the two sides ranged from 10.8% to 14.8% for all trabecular parameters except Tb.Pf (74%) and SMI (84%). In cortical bone, mAbsΔ_ind were 11.6% for Po.Dm, 15.1% for Cort.Porosity, and 27.6% for Cort.Th. To assess within-side variability, we divided the trabecular iliac crest volume into three equal parts, one adjacent to each cortex and one in the middle. Values of mAbsΔ_ind versus the middle part were ranging from 7.6% for Tb.Sp* to 26.2% for BV/TV. Thus, within-side variability was similar in magnitude to side-to-side variability. The considerable differences in robustness across trabecular parameters indicate a need for selecting the most stable parameters, most notably for longitudinal studies of small numbers of patients. Acquisition by μCT and image analysis must comply with stringent quality criteria, especially the distance from the cortices must be standardized.