Mixed effects and multivariable linear modelling of risk factors for bone mineral density loss

Thurston, Charles (2020) Mixed effects and multivariable linear modelling of risk factors for bone mineral density loss. Masters thesis, UNSPECIFIED.

Text (2020ThurstonMScbyResearch)
2020thurstonmscbyresearch.pdf - Published Version
Restricted to Repository staff only until 9 November 2025.
Download (3MB)

Abstract

Introduction Osteoporosis is a metabolic bone disease with reduced bone mineral density (BMD) that leads to increased fracture risk. Dual energy X-ray absorptiometry (DEXA) is the gold standard for osteoporosis imaging. Aims To characterise further the risk factors for BMD loss, and potential detection of novel risk factors through mixed effects modelling. Methods DEXA measurements were collected from Royal Lancaster Infirmary from 2004-2018 and patients completed a risk factors questionnaire (IRAS ID 159382). The outcome variables included the left and right hip, and the spine BMD, and explanatory variables were risk factors including FRAX risk factors. In the mixed effects modelling, an indicator function was used to account for the non-linear effect of the menopause. The dataset had longitudinal (7910 patients, 88% female, mean age 62.7 years), and baseline scans (32786 patients, 82% female, mean age 64.1 years) for multivariable linear modelling. Results Age-related BMD loss was demonstrated to different degrees within the hip. The indicator variable for menopause showed a negative association at all sites. Average percent fat showed a negative association at the hip, but a positive association at the spine. When comparing BMD between genders there was potentially a systemic bias in measurement and calculation of BMD. Discussion Age demonstrates intra-hip and intra-spine variation of BMD. This is the first study to provide a statistical method that effectively describes the menopausal population. The negative association with average percent fat at the hip is a novel finding. It has potential as a prognostic indicator and therapeutic target. This study has a larger sample size than previous studies. Mixed effects modelling allows for incorporation of age and its known effects on BMD at the population and individual level. Overall, potential new risk factors have been detected and mixed effects modelling has been shown to be a useful technique for modelling BMD loss.