Clinical implications of bone marrow adiposity identified using deep learning, phenome-wide association, and Mendelian randomization in the UK Biobank

Bone marrow adipose tissue (BMAT) is a normal feature of mammalian anatomy that increases with ageing and in osteoporosis, type 2 diabetes, and other diverse clinical contexts. However, the full scope of diseases associated with altered bone marrow adiposity remains to be determined, and whether BMAT directly contributes to human disease is unknown. To address these critical gaps in knowledge, we previously used deep learning to measure the bone marrow fat fraction (BMFF) of the femoral head, total hip, femoral diaphysis, and spine of over 44,000 participants in the UK Biobank, followed by genome-wide association meta-analyses to identify the genetic architecture of altered BMFF. Here, we use these data for phenome-wide association studies (PheWAS) to systematically investigate the diseases associated with BMFF at each site. First, we conduct a PheWAS using measured BMFF and find that it is associated with 47 incident diseases across 12 disease categories. These include not only osteoporosis, fracture, and type 2 diabetes, but also diseases not previously linked to BMAT, such as cardiovascular diseases, several cancers, and other conditions with a substantial worldwide burden on human health. Sex-stratified PheWASes further demonstrate that BMFF-associated diseases differ between males and females. We then establish polygenic risk scores (PRSs) and use PRS-PheWAS and Mendelian Randomization to explore potential causal association between BMFF and disease outcomes. These reveal that increased spine BMFF is associated with fractures but not osteoporosis, whereas genetic predisposition to increased BMFF at each femoral site is positively associated with both diseases. Intriguingly, genetic predisposition to increased spine BMFF is positively associated with type 2 diabetes, whereas this association is negative for BMFF at the total hip and diaphysis. Together, our findings substantially advance understanding of the impact of BMAT on human health and establish bone marrow adiposity as a promising biomarker and/or potential therapeutic target for improved prevention and treatment of human diseases.