Genomics-first association of pharmacogenomic risk phenotypes with adverse drug reactions in a healthcare-based population.

Variants in pharmacogenes alter drug metabolism and define actionable metabolizer phenotypes that carry an increased risk of adverse reaction upon exposure to the relevant drug.

This retrospective genomics-first analysis linked genetic and electronic health record data in 226,053 participants of the Geisinger MyCode cohort, covering 58 CPIC high-risk gene-drug associations across 11 genes. Most participants (211,920/226,053, 93.7%) had at least one actionable pharmacogenomic phenotype, and 44.4% (100,402/226,053) combined an ADR-risk phenotype with a prescription for the relevant drug. Carriers of a risk phenotype were more likely to have a documented allergy or an ADR-related medication discontinuation (OR = 1.4; P < 0.001), and 3.9% (8,913/226,053) had an adverse reaction linked to their own pharmacogenomic risk. Overall, 36,194 adverse drug reactions were recorded in 27,546 individuals (12.2% of the cohort), of which 10,719 (29.6%) occurred in carriers of a relevant phenotype. ADR incidence rose with the number of drug exposures (Pearson's correlation = 0.50; P < 0.001).

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The outcome relies on indirect EHR proxies — recorded allergies and medication discontinuations — whose attribution to genotype is not established case by case. The association is modest (OR = 1.4) and does not show that preemptive genotyping would have prevented these events, as the authors themselves concede. Supporting implementation would require a prospective comparison with adjudicated adverse reactions rather than a retrospective reconstruction from administrative fields.

Analysis by Dr Thibaut Benquey