Intrarater reliability of dual-energy X-ray absorptiometry-based measures of vertebral height in postmenopausal women

Abstract

The primary purpose was to estimate intrarater reliability of vertebral body height (VH) measures in postmenopausal women based on duplicate analyses of vertebral fracture assessment (VFA) images. The secondary purpose was to determine the consistency in classification of vertebral deformity on duplicate analyses. Thirty-two VFA were randomly selected from a database of 464 scans acquired in postmenopausal women using dual-energy X-ray absorptiometry (Discovery A; Hologic Inc., Waltham, MA). Visible endplates were marked on each image on 2 occasions (4 wk apart) by a single rater; the semiautomated software derived measures of anterior, middle, and posterior VH and classified severity of vertebral deformity. Intrarater reliability was assessed using the intraclass correlation coefficient (with 95% confidence interval [CI]) when ≥ 22 VFA could be analyzed. Reliability of grading deformity of 267 vertebrae was assessed using Cohen's unweighted kappa (with 95% CI). Reliability of anterior, middle, and posterior height measures from T8 to L4 was 0.85 and greater except for T8 anterior VH and T9 posterior VH (0.76 [0.43, 0.90] and 0.62 [0.15, 0.83], respectively). Chance-corrected agreement for 4 grades of vertebral deformity was 0.48 (0.30, 0.66) and for 2 categories (normal/mild and moderate/severe) was 0.70 (50, 0.90). Intrarater reliability was acceptable for VH measures from T10 to L4. Reliability in grading severity of deformity was improved by classifying as <25% deformity (nonfracture) and as >25% deformity (fracture).