Associations between modifiable lifestyle factors and incident fracture over ten years of follow-up: a preliminary phase for developing a fracture risk tool for the lay community — ASN Events
  1. Schedule
  2. Concurrent Session: Symposium - From Weight Loss to Bone Loss? Understanding Bone Health in the Obesity Drug Era
  3. Associations between modifiable lifestyle factors and incident fracture over ten years of follow-up: a preliminary phase for developing a fracture risk tool for the lay community

Associations between modifiable lifestyle factors and incident fracture over ten years of follow-up: a preliminary phase for developing a fracture risk tool for the lay community (#40)

Kara B Anderson 1 , Dulari H Lekamlage 1 , Lana J Williams 1 , Kara L Holloway-Kew 1 , Aminu Suleiman 1 , Julie A Pasco 1 2 3
  1. Deakin University, School of Medicine, IMPACT, The Institute for Mental and Physical Health and Clinical Translation, Geelong, Australia
  2. Barwon Health, Geelong, Victoria, Australia
  3. Department of Epidemiology and Preventative Medicine, Monash University, Melbourne, Victoria, Australia

Background/Aim

Absolute fracture risk is used to determine intervention thresholds and guide treatment. However, many factors involved in its computation are not modifiable, precluding individuals from pro-active involvement in preventative health care. This study aimed to develop a best-fit model of modifiable risk factors that could be identified and acted upon by individuals in the lay community.

Methods

This study included 1,126 women and 978 men from the Geelong Osteoporosis Study. Fragility fractures of all bones excluding skull and digits were identified by radiological review and participants followed up for 10 years unless censored by event (first fracture) or death. Missing predictor data were handled using multiple imputation by chained equations. Three variable-selection approaches were compared to identify the best-fitting Cox proportional hazards model: stepwise selection using the Akaike Information Criterion (AIC), regularized least absolute shrinkage and selection operator (LASSO) regression, and univariable screening. Candidate predictors included measured (body mass index, waist circumference, timed-up-and-go (TUG) test, gait speed, blood pressure, cholesterol, fasting blood glucose) and self-reported (medication use, mobility, self-rated health, current smoking, calcium deficiency, alcohol intake, falls, sun exposure, poor sleep, pain, anxiety, depression) variables. All analyses were conducted using R 4.4.1.

Results

The StepAIC-based model demonstrated slightly better performance (pooled C-index=0.731), with lower AIC and BIC values across all imputed datasets. The final model included age (HR:1.04, 95%CI 1.03-1.05), BMI (0.94, 0.90-0.98), male sex (0.49, 0.36-0.67), better self-rated health (0.73, 0.53-1.02), oral glucocorticoid use (2.10, 0.98-4.51), antifracture medication use (1.62, 0.99-2.66), summer sun exposure (1.28, 0.99-1.65), high density lipoprotein cholesterol (1.68, 1.16-2.46), waist circumference (1.02, 1.00-1.04), gait speed (1.45, 0.91-2.29) and TUG (1.03, 1.00-1.05).

Conclusions

These data suggest that individuals may be able to reduce their fracture risk, even if only marginally, through careful consideration of summer sun exposure, medication use, cholesterol levels, waist circumference and mobility. Future work involves validating the findings and refining the tool for ease of interpretation and community use.