Limited clinical utility of a genetic risk score for the prediction of fracture risk in elderly subjects

Joel Eriksson, Daniel S. Evans, Carrie M. Nielson, Jian Shen, Priya Srikanth, Marc Hochberg, Shannon McWeeney, Peggy M. Cawthon, Beth Wilmot, Joseph Zmuda, Greg Tranah, Daniel B. Mirel, Sashi Challa, Michael Mooney, Andrew Crenshaw, Magnus Karlsson, Dan Mellström, Liesbeth Vandenput, Eric Orwoll, Claes Ohlsson

Research output: Contribution to journalArticlepeer-review

31 Scopus citations


It is important to identify the patients at highest risk of fractures. A recent large-scale meta-Analysis identified 63 autosomal singlenucleotide polymorphisms (SNPs) associated with bone mineral density (BMD), of which 16 were also associated with fracture risk. Based on these findings, two genetic risk scores (GRS63 and GRS16) were developed. Our aim was to determine the clinical usefulness of these GRSs for the prediction of BMD, BMD change, and fracture risk in elderly subjects. We studied two male (Osteoporotic Fractures in Men Study [MrOS] US, MrOS Sweden) and one female (Study of Osteoporotic Fractures [SOF]) large prospective cohorts of older subjects, looking at BMD, BMD change, and radiographically and/or medically confirmed incident fractures (8067 subjects, 2185 incident nonvertebral or vertebral fractures). GRS63 was associated with BMD (≅3% of the variation explained) but not with BMD change. Both GRS63 and GRS16 were associated with fractures. After BMD adjustment, the effect sizes for these associations were substantially reduced. Similar results were found using an unweighted GRS63 and an unweighted GRS16 compared with those found using the corresponding weighted risk scores. Only minor improvements in C-statistics (AUC) for fractures were found when the GRSs were added to a base model (age, weight, and height), and no significant improvements in C-statistics were found when they were added to a model further adjusted for BMD. Net reclassification improvements with the addition of the GRSs to a base model were modest and substantially attenuated in BMD-Adjusted models. GRS63 is associated with BMD, but not BMD change, suggesting that the genetic determinants of BMD differ from those of BMD change. When BMD is known, the clinical utility of the two GRSs for fracture prediction is limited in elderly subjects.

Original languageEnglish (US)
Pages (from-to)184-194
Number of pages11
JournalJournal of Bone and Mineral Research
Issue number1
StatePublished - Jan 1 2015


  • Dxa
  • Fracture risk assessment
  • General population studies
  • Human association studies
  • Osteoporosis

ASJC Scopus subject areas

  • Endocrinology, Diabetes and Metabolism
  • Orthopedics and Sports Medicine


Dive into the research topics of 'Limited clinical utility of a genetic risk score for the prediction of fracture risk in elderly subjects'. Together they form a unique fingerprint.

Cite this