Genomic Testing is Best Integrated into Clinical Practice When it is Actionable

A fundamental principle in the judicious practice of medicine is to order only those tests whose results will guide clinical management. Such reliance on the actionability of medical information is consistent with the goals of minimizing the potential harms and wasted healthcare resources of unnecessary or low-value testing. Genomic medicine has been defined as a “medical discipline that involves using genomic information about an individual as part of their clinical care (e.g. for diagnostic or therapeutic decision-making) and the health outcomes and policy implications of that clinical use.”[1] This emerging field is subject to the same considerations as the rest of medicine, as clinicians, health systems, and payers apply principles of medical stewardship to the ever-changing landscape of genomic testing. These end-users will generally not be persuaded to adopt the widespread implementation of genomic testing until some threshold of actionability is achieved. And yet, the actionability of a given genomic test is not necessarily an objective determination but instead relies on a less-defined method of evaluation. This, in part, relates to the complexity of patient management, including considerations of clinical history and family history, making genomic information something that clinicians must factor into management decisions. Here, we illustrate the subjective and situation-dependent nature of the clinical actionability of genomic information in three key contexts: (1) guiding pharmacotherapy, (2) measuring disease risk, and (3) determining familial risk. By describing these contexts, we call for broader awareness and recognition of the subjective nature of genomic testing to facilitate its implementation where clinically appropriate.

Guiding pharmacotherapy

Perhaps the readiest examples of the actionability of genomic information are its applications in guiding treatment with medications. Pharmacogenomics encompasses the use of an individual’s distinct genomic profile to facilitate more efficacious or safer medication choice and dosing. The United States Food and Drug Administration labels for more than 200 medications include information about how patient genotype might impact drug action, and the Clinical Pharmacogenetics Implementation Consortium has found sufficient high-quality evidence to support the development of drug dosing recommendations for dozens of drug-gene pairs.[2,3] The use of genomic information has already significantly impacted treatment decisions for many patients with cancer. An increasing number of drugs are co-developed with companion genomic diagnostics in recent years, including vemurafenib for melanoma patients with the BRAF V600E mutation and crizotinib for non-small cell lung cancer patients who are ALK positive.[4]

Clinicians value genomic information with clear actionability, and yet even the value of pharmacogenomic testing is context-specific. In a qualitative study of provider clinical decision-making around pharmacogenomics, primary care providers stated that pharmacogenomics has value for specific conditions or disease categories, namely, psychiatry, cardiology, oncology and pain management.[5] We have recently described how a large healthcare system might choose to adopt certain pharmacogenomic tests and not others, based on considerations such as therapeutic alternatives or insufficient evidence of improved outcomes, despite similar validity of the gene-drug associations in question.[6] Moreover, a pharmacogenomic test result clearly has no value if the patient has no clinical indications for the associated medication. Similarly, if a pharmacogenomic test result is obtained decades before it is clinically relevant for a patient, in today’s healthcare environment and disjointed medical record infrastructure it is unlikely that the result will be readily available to support timely medical decision-making.

Measuring disease risk

Genomic information is also used to determine disease risk and identify subsequent interventions. Genomic risk profiling can be based on either a single variant, as with a pathogenic BRCA1 variant and hereditary breast and ovarian cancer, or millions of variants across the genome, as in a polygenic risk score for coronary artery disease (CAD).[7,8] The actionability of such genomic risk assessment depends on myriad factors including the clinical and family history of the patient. While there is U.S. Preventive Services Task Force consensus that a well-described pathogenic variant in BRCA1 should be a call to clinical action in the form of heightened breast cancer surveillance or prophylactic mastectomy, the actionability of a polygenic risk score for CAD remains unclear, particularly in the context of current risk stratification practices based on non-genetic factors.[8]

The Clinical Genome Resource (ClinGen) Actionability Working Group has recently defined actionability more objectively, proposing a semi-quantitative metric comprised of four qualitative domains : disease severity, likelihood of disease, effectiveness of an intervention, and nature of the intervention. This semi-quantitative metric has now been applied to more than 100 pairs of monogenic diseases. For instance, the ClinGen actionability curation or summary for BRCA1-related breast and ovarian cancer suggests that there is definitive genetic and experimental evidence to support the claim that variations in BRCA1 are associated with breast and ovarian cancer.

Recognizing variation among children in the typical age at disease onset and the typical age at disease intervention, Milko recently proposed an age-based semiquantitative metric for use in selecting monogenic conditions for inclusion in exome and genome sequencing for newborn screening. This metric would assist parents and physicians in making informed decisions about the disclosure and actionability of test results.[9] Such systematic approaches to the evaluation of the clinical actionability of polygenic risk scores could help build consensus around their readiness for implementation in various clinical contexts.

Determining familial risk

Genomic testing can also be used to determine familial disease risk. Whether these implications for family members make genomic testing medically actionable for the patient has classically been contested but more recently emerged as a type of extended clinical utility.[10] A recent qualitative study assessed clinical genetics experts’ perspectives on the clinical utility of genomic testing; half identified family impact as a key domain.[11] A pediatric geneticist among the interviewees shared the opinion that even if genetic testing would not directly alter patient management, cascade testing in family members would aid in decision-making for prenatal testing and family planning. A separate study of patient perspectives identified similar constructs, finding that testing can help families understand possible health risks for themselves and their future children, providing the families with assurance in the form of a more definite diagnosis or clearer prognosis for both current and future generations.[12] An example is alpha-1 antitrypsin deficiency, a genetic condition that causes early-onset pulmonary emphysema, where high value is placed by patients and their families on knowledge gained from genomic testing and counseling to facilitate risk modification and monitoring.[13]

Genomic testing to determine familial risk is also context-dependent; it depends on factors such as numbers and ages of living relatives, plans for future family planning, preferences among family members to receive and share genomic information, and the penetrance of the disease in question. One recent study assessed the psychosocial effects of risk prediction in families with newborns determined to have an elevated genetic risk of type 1 diabetes.[14] None of the children in the cohort went on to develop diabetes over the next 12 years, and their parents expressed that although they were not psychologically burdened by knowledge of their children’s risk, they perceived little benefit from knowing this information. Overall, the parents were unenthusiastic about expanding genomic testing into newborn screening, although most disclosed this genomic risk information to their children by 12 years of age without any obvious adverse effects.[14] Thus, given that genomic testing has the potential to impact family relationships or cause psychological impacts, providers are unlikely to implement genomic testing for diseases like type 1 diabetes in children if parents see little value in the information for their children.

Discussion

We have presented three broad categories to illustrate how the clinical actionability of genomic testing is not a simple yes/no determination but, rather, lies on a continuum influenced by the natural history and medical management of the condition in question and context-dependent factors of the end-users. Genomic testing is the fastest growing sector of medicine. As of August 2019, there were 60,029 genetic tests for 11,982 conditions and 18,630 genes in the National Center for Biotechnology Information Genetic Testing Registry.[15] Genomic testing holds significant and growing importance in medicine.

It is also important to consider how the clinical implementation of various genomic testing strategies are influenced by provider- and health system-level factors. Even if a genetic test result is actionable, that fact alone will not promote its implementation if other provider- and systems-level factors are not in place to facilitate its use and act upon the results. We agree with the previous opinion of Ginsburg that the current implementation science models should be adapted to meet the context-specific needs of patients, providers and health systems.[16] Stakeholders and health systems seeking a threshold of actionability prior to the widespread promotion and implementation of genomic testing should consider the subjective and context-specific nature of actionability that we have illustrated here, in addition to ongoing efforts to measure that actionability more objectively.