Limitations of pharmacogenomic data in FDA-approved drug labels

Published: October 15th, 2014

Category: Stories

Pharmacogenomic biomarker information is included in Food and Drug Administration (FDA)-approved product labeling, or prescribing information, for over 100 drugs. This information may be used to support the clinical use of pharmacogenomic data to optimize drug therapy; however, there is wide variation in the scope and format of FDA-approved labeling. An analysis of pharmacogenomic information included in FDA labeling was recently published in JAMA Internal Medicine.

Wang and colleagues1 investigated pharmacogenomic biomarker data available in publicly accessible FDA drug databases. Authors reviewed and categorized evidence included in the drug label according to guidelines from the Evaluation of Genomic Applications in Practice and Prevention (EGAPP) Working Group in regards to clinical validity (i.e., ability of test to predict phenotype) and clinical utility (i.e., likelihood that test will improve clinical outcomes). Evidence was classified as being convincing, adequate, or incomplete. Additional supporting evidence was included only when specifically cited in the drug label. Evidence was rated as incomplete if the study design was not adequately described and cited in the package insert.

Study investigators identified 119 drug–biomarker combinations, with 63% of recommendations related to adverse events and 37% pertaining to drug efficacy. Authors reported that 43 drug labels (36.1%) provided convincing evidence of clinical validity and 18 labels (15.1%) demonstrated evidence of clinical utility. Biomarkers for oncology drugs were more likely to have evidence of clinical utility than other drugs (14 of 37 [37.8%] vs. 4 of 82 [4.9%], P <0.001).

One-half (17 of 34) of drug labels for targeted therapies (i.e., medication developed to interfere with or manipulate a specific biomarker) demonstrated evidence supporting clinical utility, while labeling for only one out of 85 (i.e., abacavir) non-targeted therapies showed evidence of clinical utility. A total of 61 labels (51.3%) included recommendations about clinical decision making, with 28.6%  of labels basing recommendations on the drug’s mechanism of action and 22.7% on drug–biomarker associations.

Clinical Implications

Findings of the current study provide insight about biomarker data included in FDA labeling and reinforce known deficiencies in the FDA’s process of developing dynamic and up-to-date prescribing information. However, study findings should be interpreted with caution given some important limitations.

First, authors highlighted differences in supporting evidence based on “targeted” versus “non-targeted” therapies. They did not, however, acknowledge other factors that influence these differences, such as whether evidence was generated in a pre-approval or post-marketing environment or was intended to support clinical decisions in regards to efficacy or safety. FDA has stated that expectations for evidence supporting the approval of a drug that requires a pharmacogenomic test for appropriate use differ from those of evidence informing safety decisions in a post-marketing environment.2,3 In the latter case, retrospective or pharmacokinetic data are often used to support biomarker- and other safety-related post-marketing label changes in an iterative process to keep prescribers informed of emerging evidence that influences clinical decision making. With this in mind, it is logical and expected that non-targeted therapies would be most often associated with the type of retrospective data that do not meet clinical utility thresholds as defined by investigators of the current study.

Investigators also failed to acknowledge that challenges inherent in maintaining a product labeling system which is dynamic and responsive to emerging scientific literature are not unique to pharmacogenomics. Similar inconsistencies and heterogeneity regarding the application of current evidence have been well documented within product labeling for drug-induced hepatotoxicity, renal dosing adjustments, drug–food interactions, and drug–drug interactions in the United States and many other countries.4-7

Given the known challenges of maintaining current and accurate product labeling from a regulatory perspective, authors were accurate in their call for physicians to investigate other evidence sources, including the primary literature. Prescribers frequently rely on data sources other than the package insert in clinical decision making, including drug databases, published information in guidelines, journal articles, and others.8,9 However, authors of the current study failed to acknowledge the role of existing evidence from these sources within pharmacogenomics, most notably the emerging data available through the Pharmacogenomics KnowledgeBase (PharmGKB), their categorized information on FDA drug labels, and existing clinical guidelines from the Clinical Pharmacogenetics Implementation Consortium. PharmGKB’s curated database of emerging literature has been cited as a source of evidence that is on par with the FDA’s publicly available data used by study investigators.10 A recent analysis of physicians’ educational needs in the area of pharmacogenomics concluded that efforts to increase physician awareness of PharmGKB should be undertaken to help inform appropriate clinical decision making.11

The authors of a commentary accompanying this study12 appropriately concluded that FDA drug labels unfortunately add to the confusion on how to interpret and apply the evidence surrounding pharmacogenomic testing in a clinical environment. It is also important in applying study findings, though, to differentiate between the approved wording in the FDA label and the body of available evidence supporting pharmacogenomic testing: they are not one and the same.


  1. Wang B et al. Clinical evidence supporting pharmacogenomics biomarker testing provided in US Food and Drug Administration labeling. JAMA Inter Med. 2014 Oct 13 [Epub ahead of print].
  2. Surh LC et al. Learning from product labels and label changes: how to build pharmacogenomics into drug-development programs. Pharmacogenomics. 2010;11:1637–47.
  3. Lesko L et al. DNA, drugs, and chariots: on a decade of pharmacogenomics at the US FDA. Pharmacogenomics. 2010;11:507-12.
  4. Boyce RD et al. Dynamic enhancement of drug product labels to support drug safety, efficacy, and effictiveness. J Biomed Sem. 2013;4:5.
  5. Pfistermeister B et al. Inconsistencies and misleading information in approved prescribing information from three major drug markets. Clin Pharmacol Ther. 2014;96:616-24.
  6. Willy ME et al. What is prescription labeling communicating to doctors about hepatotoxic drugs? A study of FDA approved product labeling. Pharmacoepidemiol Drug Saf. 2004;13:201-6.
  7. Seminario MJ et al. Are drug labels static or dynamic? Clin Pharmacol Ther. 2013;94:302-4.
  8. Vaughan KTL et al. An evaluation of pharmacogenomic information provided by five common drug information sources. J Med Libr Assoc. 2014;102:47-51.
  9. Stanek E et al. Adoption of pharmacogenomic testing by US physicians: results of a nationwide survey. Clin Pharmacol Ther. 2012;91:450-8.
  10. Tutton R. Pharmacogenomic biomarkers in drug labels: what do they tell us? Pharmacogenomics. 2014;15:297-304.
  11. Johansen Taber KA et al. Pharmacogenomic knowledge gaps and educational resource needs among physicians in selected specialties. Pharmgenomics Pers Med. 2014;10:145-62.
  12. Burke W et al. A call for accurate pharmacogenetic labeling: telling it like it is [commentary]. JAMA Inter Med. 2014 Oct [Epub ahead of print].

Tagged as: ,