Near, far, wherever you are: Phenotype-related variation in pharmacogenomic effect sizes across the psychiatric drug literature.

Abstract

BackgroundPharmacogenomics is viewed as one route to minimising inter-individual variability in drug response. However, uptake in psychiatry is slower than in other medical fields, so assessing evidence for psychiatric genotype-drug pairs and understanding what influences the magnitude of these effects is essential. MethodsWe performed a systematic search for studies investigating pharmacogenomic variation in the context of antipsychotic and antidepressant use. Outcomes varied, including those related to drug bioavailability ("proximal") or side effects, symptom severity, and other treatment outcomes ("distal"). We performed a meta-analysis moderated by outcome type to quantify the average pharmacogenomic effect size across proximal and distal outcomes and assess whether they differ from one another. We developed a companion dashboard that allows users to explore the dataset and perform simplified meta-analyses, power calculations, and Bayesian shrinkage analyses based on drugs, enzymes, and outcomes of interest. ResultsWe analysed 2,092 standardised mean differences (SMDs) from 184 studies, finding evidence that pharmacogenomic effect sizes for proximal outcomes were significantly larger than distal ({Delta}{beta} = -0.206 [95% CI -0.291 - -0.121], p = 5x10-6). This trend was consistent across sub-groups restricted to the most common gene-drug pairings in the dataset. Power calculations for hypothetical future studies using two-sample t-tests showed that, to attain at least 80% statistical power, analyses of distal outcomes require a larger sample size than proximal outcomes. DiscussionWe demonstrate that pharmacogenomic effect sizes are significantly larger for proximal outcomes related to pharmacokinetics than for distal outcomes related to efficacy and toxicity. Understanding how the biological mechanisms underlying different outcomes might impact pharmacogenomic effect sizes could help to inform participant recruitment for future psychiatric pharmacogenomic studies, alongside the development of pharmacogenomic guidelines for psychiatric medications. FundingThis work was supported by a PhD studentship from Mental Health Research UK (SKL). Authors were also supported by grants from the Medical Research Council (SEL, MCOD, AFP; MR/Y004094/1 and MR/Z503745/1), the European Unions Horizon 2020 programme (DBK, MCOD, AFP; #964874), the European Unions Horizon 2021 programme (AFP, #101057454), and the Dementia Research Institute (ES; UKDRI supported by the Medical Research Council (UKDRI-3206), Alzheimers Research UK, and Alzheimers Society). The authors declare no conflict of interests.