Retrospective Analysis of Factors Associated with Fracture in 714 Patients with Polymyalgia Rheumatica

Abstract

Introduction. Polymyalgia rheumatica (PMR) is a disease of the elderly, associated with increased fracture risk due to glucocorticosteroid (GC) treatment with the additional possible influence of chronic inflammation. Risk factors for fracture in PMR have not been extensively studied. Hip structure analysis (HSA) is a way to measure bone morphology in the hip using dual X-ray absorptiometry (DEXA). It has been used as a predictor of fracture in epidemiological settings. HSA has not been studied in PMR before. Objectives. The object of this retrospective study was to determine if fracture risk in PMR was associated with densitometry data and to determine the influence, if any, of HSA on that association. Methods. 714 patients with PMR referred for a bone density estimate at a district general hospital from June 2004 to October 2010 were studied. Demographic data, GC use, alcohol consumption, smoking status, secondary osteoporosis, and fracture history were recorded. Bone mineral density (BMD), Z score, T score, body composition data, and HSA measurements were collected. These were geometric measurements taken from 2-dimensional DEXA images of the hip. Fracture was modelled as an outcome variable using logistic regression models, adjusted for age and sex. And the fit of the model was assessed by comparing the area under the curve (AUC). Results. 714 patients were studied, 532 (75%) were female, and mean age was 70.5 with SD of 8.8. 703 (98%) had been treated with GCs. Lumbar and femoral BMD models were significantly associated with fracture. Right femur OR 0.062 (0.014-0.285), left femur OR 0.098 (0.023-0.412), right femoral neck 0.078 (0.014-0.43), left femoral neck 0.104 (0.022-0.492), L1 0.192 (0.066-0.56), L2 OR 0.138 (0.053-0.358), L3 0.192 (0.079-0.463), and L4 0.243 (0.108-0.544). Cross-sectional area was the only HSA parameter that was associated with fracture OR 0.988 (0.980–0.997). Conclusion. L2 association models were strongest. Prospective studies are needed to elucidate whether these factors predict future fracture. GC data were binary, not reflecting dose and duration.