Does hip shape associate with hip pain and early structural and radiological changes at the hip in an Australian community-based sample?

Abstract

Purpose: Study of bone morphologic features of the hip can be used to predict hip fractures, total hip replacement (THR) and radiological hip OA (ROA). However, limited data exists on this concept. Here we describe associations between hip shape and features of hip OA in a large community based sample. Methods: Baseline dual energy X-ray absorptiometry (DXA) images of left hip of 831 subjects from the Tasmanian Older Adult Cohort (TASOAC) were included. An 85-point hip shape model was generated from the DXA images using the active shape modeling (ASM) toolkit (Manchester University UK) and Shape software (University of Aberdeen, UK). Modes 1-6, in decreasing order of variance in modes were extracted. Within the dataset each mode was normalized and is independent of other modes; thus every mode (from 1-6) is an independent descriptor of hip shape. Hip pain was assessed by WOMAC. Lower limb muscle strength was measured by dynamometer and hip fracture was self reported. Presence of hip cartilage defects, hip bone marrow lesions (BMLs), hip cartilage volume and hip effusion cross-sectional area (CSA) were assessed using MRI. Joint space narrowing (JSN, 0-3) and osteophytes (0-3) were assessed on x-ray using Altman’s atlas. Log binomial regression (STATA 12, Texas, USA) was used to estimate the associations of hip shape and data were adjusted for age, sex and BMI accordingly. Results: The first six shape modes extracted from the dataset, describe 68% of shape variation in this study sample. In comparison to the rest of the modes, characteristic differences were most prominent for subjects with modes 1 and 2 shape variation. Shape mode 1 was associated with leg strength (PR:3.66 95%CI:1.40-5.94), presence of effusion (PR:1.28 95%CI:1.02-1.62), hip cartilage volume (beta:263.3 95%CI:105.3-421.4). Mode 2 did not associate with presence of hip pain but with worsening hip pain in those with greater BMI (PR:3.50 95%CI:0.20-6.79). Mode 3 did not show any associations. Mode 4 was associated with lower prevalence of both hip fractures (PR:0.40 95%CI: 0.20-0.96) and presence of hip BMLs (PR:0.63 95%CI:0.42-0.93). Mode 5 was not associated with any structural or radiological features while mode 6 associated with higher prevalence of ROA (PR:1.14 95%CI 1.05-1.30). Conclusions: Each mode describes independent attributes of the shape and can be used to study specific features of hip OA. In this sample, shape of the hip is associatedwith clinical, structural and radiological features. Moreover,modes 1 and 2 in thismodel describe 45% of the total variance and are associatedwithmeasures of bone, cartilage andmuscle andwith worsening pain. Mode 6 was a good descriptor of radiographic OA.