This study examines the influence of iterative reconstruction on bone mineral density (BMD) measurement by comparison with standard quantitative computed tomography (QCT; reference) and two other protocols based on filtered back projection. Ten human cadaver specimens of the lumbar spine with a hydroxyapatite calibration phantom underneath, were scanned with 4 protocols: 1. standard QCT, 2. volume scan with FBP, 3. helical scan with FBP, and 4. helical scan with IR (Adaptive Iterative Dose Reduction 3D (AIDR3D)). Radiation doses were recorded as CT dose index (CTDIvol) and BMD, signal-to-noise and contrast-to-noise ratio were calculated. Mean hydroxyapatite concentration (HOA) did not differ significantly between protocols, ranging from 98.58 ± 31.09 mg cm3 (protocol 4) to 100.47 ± 30.82 mg cm3 (protocol 2). Paired sample correlations of HOA values for protocol 4 and protocols 1, 2 and 3 were nearly perfect with coefficients of 0.980, 0.979 and 0.982, respectively (p < 0.004). CTDIvol were 7.50, 5.00, 6.82 (±2.03) and 1.72 (±0.50) mGy for protocols 1, 2, 3 and 4 respectively. Objective image quality was highest for protocol 4. The use of IR for BMD assessment significantly lowers radiation exposure compared to standard QCT and protocols with FBP while not degrading BMD measurement.