Concordance of Drug Labels and Clinical Annotations

Some drug labels have been identified by the FDA as containing pharmacogenetic biomarker information, but occasionally these variants are not listed in the Clinical Annotations on PharmGKB. The Clinical Annotations are based on literature available in peer reviewed journals and focus on germline variants. As a result, there are a few reasons why variants identified in drug labels may not be found in PharmGKB Clinical Annotations.

1. Data used to create the drug labels are not publicly available. Research conducted by pharmaceutical companies, such as trials done for drug approval, may be proprietary information and/or not be published, but are used by the FDA. Only published literature is curated in PharmGKB.  

2. FDA labels may be based on drug classes but Clinical Annotations are drug specific. For example, the label of Protriptyline contains information about CYP2D6 but it is unknown whether proprietary information or drug class generalizations were used for this associations. Protriptyline is a tricyclic antidepressant. While no literature is available in a PubMed search for pharmacogenetics of protriptyline itself, the protriptyline drug label has a precaution for all tricyclic antidepressants, warning that poor metabolizers via CYP2D6 have higher plasma concentrations of tricyclic antidepressants generally.

3. Data are based on studies that do not identify specific variants or that are done on functional protein assays. Clinical Annotations are written about specific SNPs or haplotypes, not for genes or pathways generally. For example, Tetrabenazine is used to treat hyperkinetic movement disorder, such as in Huntington’s Disease, and CYP2D6 testing is required by the FDA. The study describing altered metabolism of tetrabenazine based on CYP2D6 metabolizer status (Mehanna, 2013) classifies CYP2D6 metabolizer status by phenotypes but does not include any specific genotyping information. As a result, there are no variants to annotate in the case of tetrabenazine, though altered CYP2D6 activity has been associated with tetrabenazine response. Another example is valproic acid, which is contraindicated in patients with Urea Cycle Disorders (UCD). Therefore, genetic variation in any of the genes involved in the urea cycle is actionable, but not specifically annotatable.

4. Relevant variation is in tumor cells and not germline. Cancer drugs that target tumors may be active or inactive based on mutations in the tumor cells. These drugs include afatinib, which is used to treat non-small cell lung cancer with EGFR mutations. Because these mutations occur in the tumor cells and are not part of the germline DNA, they are not covered extensively in PharmGKB, though we are working on ways to expand coverage.

5. PharmGKB does not have the resources to cover all the articles of PubMed, and some annotations may not be curated yet. For routine curation, we focus on journals with a high volume of pharmacogenetics articles. Articles published in other journals may lag behind in curation. If you know articles that you think should be added to curation, please send them to feedback@pharmgkb.org.