Comparing morphometric X-ray absorptiometry and radiography in defining vertebral wedge fractures in patients with ankylosing spondylitis

Abstract

To compare the level of agreement of quantitative morphometry of the vertebrae on lateral views of the spine using conventional X-ray and using a dual X-ray absorptiometry device (DXA) in determining the degree of wedging of vertebrae in patients with ankylosing spondylitis (AS). Thirty patients with AS underwent DXA to acquire single-energy morphometric X-ray absorptiometry (MXA) scans and conventional lateral radiography (MRX) of the thoracic and lumbar spine. Vertebral anterior and posterior heights were measured and the anterior/posterior (AP)-ratio was calculated. We analysed the level of agreement for vertebral wedging between MRX and MXA on the patient level and on the vertebral level, using average AP-ratios per patient, and per vertebra, as well as dichotomized AP-ratios (above or below cut-off levels that are commonly used to identify fractures). Per-patient analysis showed good agreement between both methods in the whole spine [intraclass correlation coefficient (ICC) = 0.64], as well as in the thoracic (ICC = 0.66) and lumbar spine (ICC = 0.62) separately. Analysis on individual vertebrae showed differences in agreement dependent on which part of the spine was measured. The ICC on all vertebrae was 0.71, 0.76 in the lumbar and 0.43 in the thoracic vertebrae. If AP-ratios were translated into fractures (yes vs no) using different cut off levels for a fracture (AP-ratios 0.75, 0.80 or 0.85) between-method agreement became fair to good (kappa 0.26-0.35 in the thoracic and 0.47-0.80 in the lumbar vertebrae). Differences in classifications were in both directions and in all vertebral fractures according to the Genant definition. In this study with a prevalence of 5% of vertebral fractures, the positive predicted value (PPV) was 39% and the negative predicted value (NPV) was 97%. Although the agreement between MRX and MXA in measuring global vertebral wedging, expressed as (mean) AP-ratio, was good, the reliability of both measures to assess wedging at the vertebral level was highly variable, ranging from fair to very good agreement, dependent on the level. If fracture studies are performed with either of both the methods, the results of wedging at the individual vertebral level cannot be generalized to the other method, except for wedging <0.75 at the lumbar spine. However, as the NPV was high, DXA could be of clinical value to select patients for further evaluation by X-ray to assess vertebral fractures as a sign of bone failure.