Machine learning augmented genome-wide meta-analysis of prescription opioid use in 860,000 individuals

Opioid analgesics are widely prescribed for pain, yet individuals show substantial variation in medical opioid use. To investigate the genetic basis of prescription-derived intake, we analyzed 859,675 Europeans across three biobanks. Prescription records were harmonized to cumulative oral morphine equivalents (OME), yielding three outcomes: any opioid prescription, cumulative dose among users, and population-level dose including non-users. Genome-wide meta-analyses identified 78, 20, and 135 loci, respectively (234 independent signals across 145 regions). All traits were highly correlated and strongly overlapped with pain-related genetics, though cumulative dose among users captured a more distinct dose-intensity component. To detect deviations from expected medical use, we trained gradient-boosted models to derive early-onset and excess-dose phenotypes. Early onset showed no genome-wide associations and mirrored pain architecture. Excess dose identified a significant signal at rs58099562 in high LD with the CYP2D6*4 loss-of-function allele and correlated more with psychiatric and substance-use traits. These results show that conventional prescription traits primarily reflect pain biology, whereas disproportionately high dosing captures distinct neuropsychiatric and pharmacokinetic liability. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=188 SRC="FIGDIR/small/25341785v2_ufig1.gif" ALT="Figure 1"> View larger version (55K): org.highwire.dtl.DTLVardef@14b003eorg.highwire.dtl.DTLVardef@fd02ceorg.highwire.dtl.DTLVardef@c33140org.highwire.dtl.DTLVardef@d674a1_HPS_FORMAT_FIGEXP M_FIG Phenotype definitions, meta-analysis, and machine learning-based refinement of opioid prescription traits. (a) Three opioid prescription phenotypes were defined: binary prescription status (RxExpPop), cumulative dose among users (RxDoseUser), and combined dosage plus binary prescription (RxDosePop). (b) Genome-wide association studies were performed across multiple biobanks and meta-analyzed using METAL. (c) Machine learning was applied to refine overuse phenotypes, resulting in early onset (RxOverUse_Onset) and high dose (RxOverUse_Amount) subtypes. (d) Downstream analyses included genetic correlation, gene annotation, biological interpretation, and cross-study comparisons. C_FIG