Abstract
Clinical PGx practice guidelines (PGx guidelines) may have limited generalizability for "marginalized" groups. We proposed the five-step Real-World Data for Genome-Guided Prescribing (ReGGRx) framework and, using All of Us research program (AoU) data, examined its ability to estimate disparities in concordance with and benefit from PGx guidelines for CYP2C19 testing when choosing antiplatelet and antidepressant drugs. The selected measures were intended to identify disparities in avoiding drug failure independent of following PGx guidelines, the odds of avoiding drug failure with PGx concordant treatment, and the degree to which "marginalized" groups (i.e., groups underrepresented in biomedical research [UBR] and with indeterminate CYP2C19 phenotypes) benefit from PGx concordant treatment, when compared with "non-marginalized" groups (i.e., non-UBR and known CYP2C19 phenotypes). Our findings identified disparities in the antidepressant cohort with UBRs (32% of cohort) having a lower odds of avoiding drug failure. For both cohorts, a lower probability of avoiding drug failure was observed in the indeterminate phenotype group (1% of cohorts) than in the known phenotype group, indicating a need to better characterize rare or ancestry-specific risk alleles. With PGx concordant treatment, negative equal opportunity difference values suggested that the UBR group was less likely to avoid drug failure than the non-UBR group. Overall, our findings illustrate the promise of the ReGGRx framework to assess PGx guideline generalizability and produce evidence for use in drug policy decisions.