On May 2nd, the American Academy of Pediatrics (AAP) released a pre-published policy statement titled “Eliminating Race-Based Medicine,” noting that race is a “historically derived social construct that has no place as a biologic proxy.” The statement provides necessary and meaningful commentary on “the medical field’s history of inaccurate applications of race correction and adjustment factors” and calls for the overdue elimination of race-based medicine. We at PharmGKB applaud this statement and believe it is important to engage with this call as pharmacogenomics professionals who urge for the implementation of personalized medicine.
Looking forward in research, there is an urgent need to direct research efforts towards underserved populations to address the issues of health disparities. Additionally, clinical implementation of pharmacogenomics needs the development of truly race-agnostic dosing guidelines and algorithms.
The terms ‘race,’ ‘ethnicity,’ and ‘ancestry’ tend to be used interchangeably when discussing a person’s origins. However, there are distinct differences in their meanings. Race and ethnicity are generally accepted to be social and cultural constructs, respectively, and are not typically assigned based on the genetic information of patients. By comparison, ancestry is a biological construct rooted in genetics.
Race or ethnicity are typically disclosed via self-report at the discretion of the patient or visual assignation by a third party (e.g. clinician, healthcare administrator) and, as such, have little grounding in genetics. The assignment of race or ethnicity also brings significant socio-economic implications which are inherent to the use of these constructs. Moreover, it is an inadequate proxy for genetic ancestry that carries additional complications for transracial adoptees and multiracial patients, among others.
PharmGKB notes that the AAP’s recommendation that “professional organizations and medical specialty societies should identify and critically examine organizational policies and practice guidelines that may incorporate race or ethnicity as independent variables or modifying factors” is a relevant and current point of consideration within pharmacogenomics.
In 2018, PharmGKB replaced the Office of Management and Budget Standards (OMB) race and ethnicity categories used in their curation efforts with a biogeographical grouping system [PMID: 30506572]. While PharmGKB serves as a global resource, these OMB groups are US-centric and, as socio-cultural measures of identity, lack the capacity to capture the scale of global human diversity. The use of these biogeographical groups is intended to standardize and to ensure consistency in communications regarding the variability of pharmacogenetic allele frequencies.
Pharmacogenomics, like other areas within genetics, is impacted by the homogeneity of reference genomes. When dosing algorithms and polygenic risk scores are developed primarily in patients of one ancestry (typically European), differences in allele frequencies across populations may prevent these tools from benefiting populations not represented in the reference data. It’s important to recognize that individual medical experiences are shaped by many different factors, including genetic, socio-economic and geographical and that the collective experience of an underrepresented ancestry or ethnic population is likely to be overlooked or diminished by a relative lack of research in these groups. A greater focus on research in underrepresented populations is crucial to understanding how medical experiences can and do vary between communities, and allows us to meet and address a wider variety of medical needs. In terms of implementation of pharmacogenetic guidance, no study population can fully characterize an individual.
Almost all CPIC guidelines are ancestry-agnostic. However, there is an exception with a complex history worth addressing, as well as a good example of the importance of expanding allele frequency research. Seminal work in warfarin pharmacogenomics focused on testing for CYP2C9*2 and *3, which are found at a frequency of 13% and 7.5% (respectively) in populations of European ancestry, with other alleles present at <1%. In populations of African ancestry, CYP2C9*2 and *3 alleles are instead found at a frequency of around 1-2% - the highest frequency reduced-function variant in this population is *8, and *5, *6, and *11 are also found at around or over 1% frequency (see the CYP2C9 allele frequency table). As such, in studies or implementation where only *2 and *3 were tested for, patients of African ancestry with these other alleles were likely not identified, and instead they were assigned a *1/*1 diplotype. This has resulted in confusing results, patients who unexpectedly required a decreased dose of warfarin, and reduced confidence in the capacity for genotyping to have the desired effect for the patients’ clinical response. Because of this, the CPIC guideline for CYP2C9, CYP4F2, VKORC1, and warfarin chose to stress the importance of testing for these alleles in populations with African ancestry.
Yet, these alleles are not absent in other populations, and this approach also fails to acknowledge that patients with mixed ancestry may not identify as of African descent and that some patients may not know whether they have African ancestry. The CPIC guideline does provide optional guidance for patients without African ancestry with CYP2C9*5, *6, *8, and/or *11 that is similar to the guidance provided for patients with African ancestry, but only for patients who already have information for these alleles - and does not include recommendations for testing for these alleles in patients without African ancestry.
Currently, these alleles do not have differing effects based on the patient’s ancestry - the distinction is in the frequency, not the clinical implementation. Testing every patient for these alleles would be beneficial, even if only to catch the few patients in other populations and those with multiple or uncertain ancestries who also carry these alleles. It is, after all, called personalized medicine.
The CPIC guideline also refers specifically to the SNP rs12777823 in recommendations for African-American patients only (not all patients of African ancestry). This is the only guideline where it states not to use a genetic result in some individuals - “the data do not suggest an association between rs12777823 genotype and warfarin dose in non-African Americans, thus rs12777823 should not be considered in these individuals (even if available).”
This SNP is found across all populations (according to gnomAD v2.1.1), and has been stated to affect warfarin response in African-American patients [PMID: 26024874] [PMID: 26877068] [PMID: 28686080]. The mechanism by which this SNP affects warfarin response is unknown - the SNP is intergenic, but close to CYP2C18, and is largely assumed to be a marker in linkage disequilibrium (LD) with something else [PMID: 23755828]. While further study is needed, it’s unlikely that this SNP itself has a function in people of one genetic ancestry but not in others, and more likely that the frequency of LD with a mechanistic variant varies by populations.
This call for race-agnostic pharmacogenetic clinical implementation is not limited to warfarin - several papers have been published discussing the U.S. Food & Drug Administration’s use of race as a limitation in recommendations for pharmacogenetic testing related to carbamazepine and allopurinol [PMID: 18840252] [PMID: 33492362]. This conversation is not a new one - from letters of support arguing against use of race to determine whether it is appropriate to test patients for HLA variants prior to initiating allopurinol therapy [PMID 30383575] to responses validating race-agnostic approaches to HLA genotyping [PMID: 30383576].
In summary, PharmGKB applauds the AAP’s statement calling for the elimination of race-based medicine. It is critical for the field of pharmacogenomics to examine how we can incorporate race-agnostic approaches to PGx implementation while acknowledging the critical importance of research in diverse populations to identify crucial PGx variants. Race-agnostic PGx implementation has the potential to meet a wider variety of individual medical needs, particularly with respect to underrepresented populations and those patients with non-monolithic ancestry, and it continues to be critical from a scientific standpoint to further expand our understanding of allele frequencies beyond Eurocentric frames of reference.
(For those interested in learning more, ClinGen’s Ancestry and Diversity working group, in collaboration with ELSIhub, are currently running a series of conversations on populations descriptors (i.e., race, ethnicity, and ancestry), and several of the speakers have pointed to the distinctions between these terms, particularly with respect to clinical genetics.)
Teri E. Klein (Principal Investigator of PharmGKB, CPIC, PharmCAT, and Stanford ClinGen)
Michelle Whirl-Carrillo (Co-Principal Investigator and Director of PharmGKB)
Li Gong
Rachel Huddart
Ingrid Keseler
Clarissa J. Klein
Binglan Li
Caroline F. Thorn
Matt W. Wright (Director, Stanford ClinGen)
Mark Woon
Michelle Whirl-Carrillo (Co-Principal Investigator and Director of PharmGKB)
Li Gong
Rachel Huddart
Ingrid Keseler
Clarissa J. Klein
Binglan Li
Caroline F. Thorn
Matt W. Wright (Director, Stanford ClinGen)
Mark Woon
