A statistical model of shape and density of the proximal femur in relation to radiological and clinical OA of the hip

SummaryObjectiveCommon radiological measures of osteoarthritis (OA) relate poorly to symptoms as experienced by patients. We created a statistical model of shape and density to see if Dual Energy X-ray (DXA) images of the hip contain symptom-related information that is not captured by common radiological measures.MethodsDXA images of the hip were made in a prospective study of patients that met the American College of Rheumatology (ACR) criteria for hip OA. From the DXA scans, we constructed a statistical model of the appearance (shape combined with density) of the proximal femur of left and right side. The model yields a number of independent descriptors of the appearance (modes) which we related to various measures of radiological and clinical OA. These outcome measures were defined using Joint Space Width (JSW), Kellgren and Lawrence (K–L) scores, Visual Analogue Scale (VAS) and Western Ontario MacMaster Universities (WOMAC) pain scores and a self-reported global assessment score.ResultsVarious modes showed significant relations with measures of OA. Interestingly, the modes that related well with radiological OA did not relate to clinical OA and vice-versa. Moreover, the modes were predictors of status and progression of clinical OA, independent from JSW and K–L.ConclusionStatistical modeling of the appearance captures the patterns of variation in projected femoral morphology as visible on DXA images. We showed that these descriptors of subtle aspects of shape and density of the hip contain information about clinical status which common radiological measures do not. The presented results warrant further careful study of the method as a monitoring tool in clinical trials.