The proxy measure nearly all bone genetics research actually uses, from a single study of 142,487 people that went unusually far to check its statistical findings against real mouse biology — including 120 knockout mice.
Solid lines are connections this site curates. Dashed lines mean the two ends share a research paper — worth knowing, and not a claim that one explains the other.
Bone mineral density, estimated here by quantitative ultrasound of the heel, is what almost all genetic research on bone health actually measures — because it is a number obtainable from a quick scan on thousands of healthy people, unlike waiting to see who eventually breaks a bone. This site's own osteoporosis page makes a point of noting that distinction directly, and highlights one unusual study that measured fracture itself instead. This page is the proxy measure's own honest counterpart: a single, very large study of it, on its own terms.
Kemp et al. 2017 ran a genome-wide association study in 142,487 people from UK Biobank, finding 307 conditionally independent variants at 203 loci — 153 of them previously unreported — together explaining roughly 12% of the variation in heel bone mineral density. Several were rare variants with unusually large effects.
What sets this study apart is what it did after finding the statistical associations. Rather than stopping at a locus list, the researchers ran skeletal phenotyping on 120 knockout mice, each missing a gene near one of the lead signals, and looked directly at their bones. Abnormal skeletal phenotypes turned up tied to a further 100 prioritised genes — and the analysis identified GPC6 as a genuinely novel determinant of bone mineral density, a finding statistics alone could suggest but not confirm.
Heel ultrasound is not the only proxy for bone density — total-body density from a DXA scan is another, and the two are correlated but not identical (genetic correlation 0.57 in the analysis below). Pei et al. 2019 combined the same 142,487-person heel BMD scan above with a separate meta-analysis of 30 studies measuring total-body BMD in 66,628 people from the GEFOS Consortium, using a statistical method built for exactly this situation — multi-trait analysis of GWAS (MTAG) — to borrow power across both traits at once. The joint analysis found 18 new loci that neither study reached on its own; 6 of them are on this page.
2025-10-01 · Genome-wide Association Studies of over 30,000 Samples with Bone Mineral Density at Multiple Skeletal Sites and Its Clinical Relevance. Genomics, Proteomics & Bioinformatics. 2025. DOI:10.1093/gpbjnl/qzaf097
Bone density at the skull, unexpectedly, is genetically linked to brain aneurysm risk
This study explicitly set out to move GWAS findings toward clinical use, analyzing DXA-derived bone mineral density (BMD) at 11 skeletal sites in over 30,000 UK Biobank Europeans. It found 91 independent loci across the BMD traits and fracture risk, including 5 genuinely novel ones (ABCA1, CHSY1, CYP24A1, SWAP70, PAX1), each replicating independently in both men and women. Polygenic risk scores for BMD were independently associated with fracture risk, though combining them with the clinical FRAX fracture-risk tool only modestly improved prediction beyond FRAX alone -- an honestly reported limited clinical gain, not a breakthrough. The more unexpected finding was a genetic correlation and shared polygenicity between head-specific BMD and intracranial aneurysm (IA) risk -- these are not obviously related conditions, and this represents a genuine biological link this site's intracranial aneurysm page (which already carries a different 2026 finding) did not previously document. Integrating gene expression data, the study prioritized ESR1 and SREBF1 as druggable targets, and specifically flagged omega-3 polyunsaturated fatty acid (PUFA) supplements -- already widely available, not a new drug -- as a repurposing candidate for osteoporosis prevention worth investigating. This site's heel bone mineral density page carries 459 variants; ABCA1, CHSY1, CYP24A1, SWAP70 and PAX1 are not currently among them.
Bone health is assessed with a DXA scan, not by genotype, and fracture risk is estimated with clinical tools built from age, prior fracture, medication use and other factors — not from any variant on this page. Heel ultrasound, the measurement this study used, is a screening tool in its own right in some settings, but neither it nor the genetics behind it changes how osteoporosis is diagnosed or treated.
This page's honest place in the bigger picture is on this site's osteoporosis page: that page is built around the rarer study that measured actual hip fracture rather than a density proxy, and makes the case that a proxy measurement is not the same thing as the outcome anyone actually cares about. This page's 13 variants are real, well-powered, and even mechanistically validated in mice — and they still describe a bone-density measurement, not a fracture.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Heel Bone Mineral Density comes down to these specific, well-studied positions — not a diagnosis. 459 positions are linked to this page; the ones this page's own text discusses are shown first.
Databases, guidelines and references
No. Bone health is assessed with a DXA scan and fracture risk with clinical prediction tools, not genotype. These variants describe bone mineral density specifically, which this site's osteoporosis page explains is a proxy measure related to, but distinct from, whether a bone actually breaks.
The osteoporosis page is built around a study that measured actual hip fracture, and explicitly contrasts that with the density-proxy measurement almost all other research (including this study) relies on. Keeping them separate keeps that contrast honest rather than blurring the two together.
After finding statistical associations in 142,487 people, the researchers checked their findings against real biology: skeletal phenotyping of 120 knockout mice, which turned up abnormal bone phenotypes tied to 100 prioritised genes and identified GPC6 as a genuinely novel determinant of bone mineral density.
The 307 variants found across 203 loci in the discovery study together explain approximately 12% of the variance in heel bone mineral density — a real but partial share.
Because their own source study is a direct extension of the heel BMD scan above: Pei et al. 2019 combined the same 142,487-person heel dataset with a separate total-body BMD meta-analysis, using a method built to borrow statistical power across two correlated traits, and found 18 loci neither study reached alone.
Free to reuse. This page's text is original writing from freely-available research, licensed CC BY 4.0 — reuse it, including commercially, with attribution to MyGeneLog™. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.