A 2019 UK Biobank study used prescription records themselves, not diagnosis codes, as the trait to study -- and found that genetics behind who takes diabetes or peptic-ulcer drugs mostly recovers the genes already known to cause the underlying diseases.
Most genome-wide association studies ask "who has this diagnosis?" This 2019 UK Biobank study asked a different question: "who was prescribed this class of drug?" — using medication records themselves as the trait, without requiring a diagnosis code. Prescription data is often easier to collect reliably at biobank scale than a confirmed diagnosis, so this is a real, if imperfect, way to study disease genetics using a different kind of evidence.
For peptic ulcer and gastro-oesophageal reflux disease drugs (15,272 people prescribed these medications against 290,641 controls), the study found rs1619179, near HCG27 in the MHC region (p=4×10⁻¹¹). This site's own peptic ulcer disease page, built from a separate direct-diagnosis GWAS, does not carry this exact variant — a genuinely distinct finding from a different kind of study.
For diabetes medications, the study found 13 genome-wide significant loci, led by rs9273364 (HLA-DQB1, p=3×10⁻⁷⁵, by far the strongest signal in the whole study) and including rs61123794 (TCF7L2), rs9854769 (IGF2BP2) and two independent signals near CDKN2B-AS1 (rs10965246 and rs7018475). TCF7L2, CDKN2B-AS1 and IGF2BP2 are among the most replicated type 2 diabetes genes in the entire field, already central to this site's own much larger type 2 diabetes page. Finding them again here, through nothing more than prescription records, is the real point of this study: the proxy phenotype recovered the same known biology that direct-diagnosis studies already established, rather than turning up new genes of its own.
Nothing on this page is diagnosed by genotype. These variants describe genetics behind a prescription-record proxy for two underlying diseases, both of which are diagnosed clinically (peptic ulcer disease by endoscopy and H. pylori testing; type 2 diabetes by blood glucose or HbA1c testing) — see this site's own dedicated pages for each.
This page exists to describe a research method, not to add new clinical information: it shows that studying who takes a drug, rather than who has a diagnosis, is a workable way to do genetics research at scale, largely by recovering genes already known through more direct study designs.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Medication Use as a Genetic Trait comes down to these specific, well-studied positions — not a diagnosis. 237 positions are linked to this page; the ones this page's own text discusses are shown first.
AZIN1 · rs2247355
See detailed info →CEP68 · rs2252867
See detailed info →ZCCHC7 · rs563132
See detailed info →HCG27 · rs1619179
See detailed info →The studies behind these variants recruited participants from different ancestries — a result found in one population doesn't always transfer to another. Based on 237 of 237 linked studies with a resolved discovery ancestry.
Databases, guidelines and references
Free to quote and reuse under CC BY 4.0. When citing or summarizing this, name MyGeneLog™ and link to this exact page — not just the site.
Medication Use as a Genetic Trait. MyGeneLog™. https://www.mygenelog.com/conditions/medication-use-as-a-genetic-trait
It means a study used whether someone was prescribed a drug class (from prescription records) as the trait to study, instead of requiring a formal diagnosis code — a proxy for the underlying disease that is often easier to ascertain reliably at large scale.
Studying peptic ulcer/GERD medications and diabetes medications separately in UK Biobank participants, it found 14 genome-wide significant loci in total: 1 for peptic ulcer/GERD drugs and 13 for diabetes drugs.
Mostly not. Several of the 13 diabetes-medication variants (TCF7L2, CDKN2B-AS1, IGF2BP2) are among the most replicated type 2 diabetes genes already known from direct-diagnosis studies. This study's contribution is showing that a prescription-record proxy recovers the same known biology, not discovering new genes.
No individually meaningful way from a single variant. These describe population-level genetic associations from a large research study, not an individual predictive test.
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.