Background Osteoporosis (OP)is under recognized, under-studied and under treated in men (1). Although osteoporosis is predominant in post-menopausal women, men experience worse disability with hip fracture (2,3,). Also, men tend to experience fractures at higher bone mineral density values than women (4). Previous studies have been limited in their generalisability. The male osteoporosis risk estimation score (MORES) - included 3 variables: age; weight; and history of chronic obstructive pulmonary disease - showed excellent predictive validity. This was mainly intended for risk assessment in asymptomatic men and does not apply to men with previous fractures. This was not evaluated in the population outside USA. Neither all variables nor all age groups were included (5). Predictors of poor bone density in men have not been characterised.
Objectives To investigate the predictors of low BMD at different sites in men in observational cohort (lumbar spine and femur)
Methods A retrospective observational study using clinical database data. Men referred for DEXA scanning to the Royal Lancaster Infirmary, in North West of England (2004-2010) were included. Simple linear regression models were fitted to each measurement site in turn, using each of the known predictors for bad bone health, with a simple Bonferroni correction for multiple comparisons. A multiple regression model was fitted for all predictors. Since exploratory analyses suggested that many of the predictors were correlated with each other, all candidate predictors and first-order interactions were considered, and a stepwise backwards elimination procedure based upon the Akaike Information Criterion (AIC) was employed to select covariates.
Results 3901 men were included in the analysis. Mean age was 65. In the simple regressions, the following predictors were found to have a significant effect at at least two measurement sites: age; height; weight; Body Mass Index (BMI); X Ray evidence of osteopenia; and previous fracture(s). Body fat percentage had a significant effect at lumbar spine only, and tobacco use had a significant effect at femoral neck only. The final model included the following predictors: age; weight; body fat; Body Mass Index (BMI); previous fracture; family history of fractures; hyperparathyroidism; X Ray evidence of osteopenia; and the interaction of weight and BMI. Hyperparathyroidism was not included in the model for lumbar spine BMD. The adjusted R squared values for the models ranged from 0.16 for the lumbar spine model, to 0.28 for the femoral neck model.
Conclusions The simple and multiple regression models suggest there are different predictors of low BMD in men compared to women. The use of a sample with large numbers of patients with one or more indications, and only a small number of patients with no indications, presents a challenge for inference on the population as a whole. Given the potential value of the findings this would point to the need for a randomised sample, or the use of propensity score matching. In either case more data would need to be gathered.
Curr opin rheumatol 2008; 20(4)423-8.
Ann Fam med.2007;5(6):540-6
Disclosure of Interest None Declared