Identifying factors that predict fracture risk in polymyalgia rheumatica

Ark, Rajiv and Bukhari, Marwan (2021) Identifying factors that predict fracture risk in polymyalgia rheumatica. Rheumatology, 60 (Suppl.). i58-i59. ISSN 1462-0324

Abstract

Background/Aims  Polymyalgia rheumatica (PMR) is a chronic inflammatory condition that commonly affects the elderly and may be treated with long-term steroids. In these patients, fracture risk is higher than the general population and can lead to increased morbidity and mortality. This retrospective observational cohort study aimed to identify factors which help predict fracture risk in PMR patients. Methods  Data were collected from June 2004 to October 2010 on patients with PMR at a District General Hospital. This included dexa scan data: bone mineral density (BMD), Z score, T score, fat mass, lean mass, percentage body fat, BMI and average tissue thickness. Demographic data, steroid use, alcohol consumption, smoking, secondary osteoporosis and presence of fracture was also recorded for each patient. Fracture risk was predicted by a series of binomial logistic regression models, which were adjusted for age and sex. Odds ratios with 95% confidence intervals and area under ROC curve (AUC) were calculated. Results  714 patients with PMR were studied of whom 532 were female, the mean age was 70.5. Steroid use, secondary osteoporosis, lean mass, fat mass, BMI, average tissue thickness, average percentage fat and alcohol consumption were not significant predictors of fracture in regression models. BMD, T score and Z score predicted fracture risk. AUC of BMD was lower than that of T and Z score for each level. The AUC for L2 models were higher than other levels in BMD, T and Z score. Odds ratios, 95% confidence intervals and AUC of the significant predictors of fracture are shown in the table. Conclusion  These data suggest that BMD, T and Z score help predict fracture in PMR patients. Lifestyle factors and other body composition data from dexa scans do not predict fracture risk. Strongest predictor models were at the level of L2. FRAX could therefore underestimate the fracture risk as it uses femoral measurings. Limitations of the study are that it was retrospective and only studied patients who underwent DEXA scans. Steroid data were binary, not reflecting dose and duration of use. The study may have been underpowered to detect the impact of some factors predicting fracture risk.