With improvements in genetic technologies and the recognition that inter-individual genetic differences may affect how patients metabolize or physiologically respond to pharmacologically active substances, PGx testing has been proposed as a way to personalize medication selection or dose based on a patient’s individual genes.2 The genes encoding the CYP2C19 and CYP2D6 proteins have emerged as having potential importance in the response (therapeutic or adverse) to numerous medications.3-5 In addition combinatorial PGx tests have emerged, which find polymorphisms in a number of genes associated with pharmacologically important proteins.6,7
There is little question that such testing is now technically feasible, but for a test to be reasonable and necessary there must be sufficient evidence that it provides incremental information that changes physician management recommendations in a way that improves patient outcomes.
An abundance of research on genes and alleles (variants) has been published. To identify alleles and variants of importance, we reviewed FDA-approval documents, guidelines, and subject matter expert input.
PharmGKB is a commonly referenced resource of PGx information, particularly in following documented clinical validity and utility of relevant PGx biomarkers. PharmGKB curators create clinical annotations for gene-drug interactions and assign levels of evidence to the associations based on published evidence.8 As an example, the database contains annotations regarding 102 drugs from 3 guidelines9 for drugs relevant to psychiatry, an area where PGx testing is done routinely. CPIC is a common source of guidelines. This group has been described by Dr. Annette Taylor10:
CPIC is an NIH-funded organization with a membership of more than 300 clinicians, scientists, laboratorians, and others knowledgeable about pharmacogenetics with the purpose of facilitating the use of pharmacogenetic test results for patient care. CPIC’s goal is to address this (potential) barrier by creating freely available, peer-reviewed, evidence-based, and updatable gene/drug clinical practice guidelines. A CPIC Overview Presentation can be found at https://cpicpgx.org/resources/.
CPIC uses a rigorous and systematic system to grade levels of evidence, and only gene/drug groupings with strong evidence for actionable prescribing are selected for guideline development.
CPIC guidelines help clinicians understand how to use available genetic test results to guide prescribing.
At present, CPIC has a 23 guidelines for the dosing and administration of 46 drugs based on gene-drug interactions.11 As part of this process, CPIC reviews the clinical evidence for gene-drug interactions and assigns clinical utility “levels” reflecting their confidence in the evidentiary basis for clinicians to alter their drug administration or dosage. The CPIC levels are A, B, C, and D. The process for this determination and the definition of each category are demonstrated in the figure (attached to this LCD) and table below.12
|Level of Evidence
|Strength of Recommendation
|Genetic information should be used to change prescribing of affected drug
|Preponderance of evidence is high or moderate in favor of changing prescribing
|At least one moderate or strong action (change in prescribing) recommended
|Genetic information could be used to change prescribing of the affected drug because alternative therapies/dosing are extremely likely to be as effective and as safe as non-genetically based dosing
|Preponderance of evidence is weak with little conflicting data
|At least one optional action (change in prescribing) is recommended
|There are published studies at varying levels of evidence, some with mechanistic rationale, but no prescribing actions are recommended because (a) dosing based on genetics makes no convincing difference or (b) alternatives are unclear, possibly less effective, more toxic, or otherwise impractical or (c) few published studies or mostly weak evidence and clinical actions are unclear. Most important for genes that are subject of other CPIC guidelines or genes that are commonly included in clinical or DTC tests
|Evidence levels can vary
|No prescribing actions are recommended
|There are few published studies, clinical actions are unclear, little mechanistic basis, mostly weak evidence, or substantial conflicting data. If the genes are not widely tested for clinically, evaluations are not needed
|Evidence levels can vary
|No prescribing actions are recommended
The following are examples of genes that are commonly tested for gene-drug interactions, and evidentiary review by CPIC have resulted in a level A or B rating:
The CYP2C9 protein has clinical significance in the metabolism of several drugs including phenytoin (CPIC Level A).11,13 The recently FDA-approved drug Mayzent, which is indicated for the treatment of multiple sclerosis requires CYP2C9 genotyping for dosing in accordance with the FDA prescribing information.14 Since this LCD does not address PGx for warfarin dosing, alleles and variants relevant to warfarin are not reviewed here. The following alleles are both common and are believed to have clinical significance for phenytoin dosing: *2, *3, *5, and *6, and a joint recommendation from the Association for Molecular Pathology and the College of American Pathologists has recommended that these variants be included as part of a CYP2C9 test.15 The *1, *2, and *3 alleles are necessary to safely dose the newly FDA-approved drug Mayzent.14
CYP2D6 is a clinically important enzyme in the metabolism of a large number of medications and has CPIC level A or B gene-drug interaction and dosing guideline.11 A number of particular alleles have been reviewed as having actionable use including: *3, *4, *5, *6,*7, *10, *17, and *41,*1xN, and *2xN. Drugs that may require CYP2D6 testing for safe administration include iloperidone, clozapine, duloxetine, deutetrabenazine and valbenazine, among others.
CYP2C19 is a clinically important enzyme in the metabolism of a number of selective serotonin reuptake inhibitors as well as tricyclic antidepressants with a CPIC level A or B gene-drug interaction and dosing guideline11,13. Recently published recommendations, including a report of the Association of Molecular pathology, recommend the following alleles be included in testing as a minimum based on clinical importance and population frequency: *2,*3,*17.13,16
Several specific HLA alleles are recommended for testing including some relevant psychiatric medications.13 They include HLA-B*15:02 and HLA-A*31:01. Both of these have level A or B gene-drug recommendations from CPIC and are relevant to the use of carbamazepine and oxcarbazepine. Other relevant variants in these genes include HLA-B 57:01 and HLA-B 58:01, which are relevant to the use of abacavir and allopurinol.
For more information, the FDA has subsequently published a listing of gene-drug interactions found in drug labels that should be considered when using those medications.17
Subject Matter Panel and Contractor Advisory Committee (CAC) Meeting on June 26th, 2019
In order to get a better understanding of expert opinion of PGx testing and CPIC guidelines, a panel of subject matter experts and Carrier Advisory Committee (CAC) members from CGS, Wisconsin Physicians Services, Noridian, and Palmetto GBA was convened on June 26th, 2019 via teleconference. While only invited experts and CAC members could speak, interested members of the public who registered could listen. The full recording is also available.18 Subject matter experts on the panel included the list below. Included members may have additional titles and positions to those listed.
Mary Relling, Chair, Pharmaceutical Dept. St. Jude Children's Research Hospital
John Greden, Founder and Executive Director, University of Michigan Comprehensive Depression Center
Annette Taylor, AVP, LabCorp, Co-Business Lead, Pharmacogenomics
Stuart Scott, Associate, Associate Professor, Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai
The panel generally agreed that PGx testing has the ability to provide clinically utile information that allows treating clinicians to select and dose particular medications appropriately. PGx testing (presumably for genes associated with pharmacokinetic pathways) was described as being analogous to measuring renal function with a serum creatinine prior to dosing renally cleared medications. The panel generally agreed that single gene testing and multi- gene panels (as defined at the top of this LCD) have a role in medication dosing and selection. The panel members did not specifically recommend or support the use of any combinatorial PGx test or differentiate one over another. There was general agreement that combinatorial PGx tests with a proprietary algorithm not available for public review required independent evidence establishing their validity and utility. Additionally, a comment was made that CYP2C19 and CYP2D6 testing would most likely be the appropriate comparator in a clinical study to determine if a combinatorial PGx test provides information that improves outcomes more than single gene or multi-gene panels.
A CAC member commented that PGx testing is becoming increasingly common, and it should not be restricted by provider type.
Additional Expert Input from the CAC
In addition to the panel, a number of experts who were unable to attend provided written correspondence. These included the following:
John Logan Black, Co-Director, Personalized Genomics Laboratory, Department of Laboratory Medicine and Pathology, Mayo Clinic
To summarize, Dr. Black’s comment generally agreed with the comments of the panel. He indicated support for the use of genetics in guiding pharmacologic treatment, referencing guidelines from CPIC and the FDA. He also provided references to a number of peer-reviewed studies which have been reviewed in this LCD. As regards combinatorial testing, he noted that evidence does support their use, though he also noted that it “is unclear whether the power of combinatorial PGx is driven by a few genes or if it is absolutely due to the combinatorial effects.” While it was not discussed by the panel, a manuscript (describing the GUIDED study) submitted by Dr. Black did a retrospective comparison (using statistical modelling rather than a direct comparison) of GeneSight to single gene testing.19 This study suggested that combinatorial testing predicts poor antidepressant response and outcomes better than single gene testing. Two gene panels were not considered.
He indicated that for panel testing, he would recommend a minimum panel in psychiatry consisting of the genotypes of CYP2C9, CYP2C19, CYP2D6, HLA-A*31:01 and HLA-B*15:02.
Jose DeLeon, Professor, Psychiatry, University of Kentucky
Dr. de Leon’s comments largely agree with the panel’s comments as well, though he specifically noted the clinical utility of HLA-B*15:02 in any patient of Asian ancestry before starting carbamazepine, and for CYP2D6 and CYP2C19 for some antidepressants and some antipsychotics. Additionally, Dr. de Leon also indicated his belief that the evidence did not support the use of GeneSight. Notably, he indicated the importance of CYP2D6 and CYP2C19 and questioned the testing of other CYP genes. He also noted that the GUIDED study (reviewed above) “further demonstrated that the study results were negative and the authors had to use secondary outcomes to try to demonstrate that a negative study had positive results.”
Bruce Cohen - Director of the Program for Neuropsychiatric Research at McLean Hospital and Harvard Medical School
Dr. Cohen voiced concern for the use of PGx testing in drug selection. He noted that the FDA has published a document raising concerns about PGx testing for general use in psychiatry.20 The document notes: “…the relationship between DNA variations and the effectiveness of antidepressant medication has never been established.’’ It goes on to state as a recommendation to providers:
If you are using, or considering using, a genetic test to predict a patient's response to specific medications, be aware that for most medications, the relationship between DNA variations and the medication's effects has not been established.
However, the document also notes:
There are a limited number of cases for which at least some evidence does exist to support a correlation between a genetic variant and drug levels within the body, and this is described in the labeling of FDA cleared or approved genetic tests and FDA approved medications. The FDA authorized labels for these medical products may provide general information on how DNA variations may impact the levels of a medication in a person's body, or they may describe how genetic information can be used in determining therapeutic treatment, depending on the available evidence.18
A manuscript examining the metascience of PGx testing and providing an accompanying viewpoint reviewed 10 clinical studies of PGx in psychiatry and found that none of them were blinded and used a protocol-based comparison.21 The authors point to two evidence-based protocols that are freely available and could have been used, STAR*D and the Texas Medication Algorithm Project. As the authors state early on:
Simply put, MDD [Major Depressive Disorder] is determined by a large number of genes, and, except in rare cases, no single gene or limited gene set, even those for drug metabolism and drug targets, determines more than a few percent of the risk of illness or course of treatment.
The STAR*D study included 2,876 subjects with major depressive disorder from multiple institutions.23-24 In this study all participants started with citalopram as the initial treatment and may have advanced through additional levels of treatment up to level 4. If a subject did not respond at a given level, that subject was then advanced to the next level of treatment, which included alternative treatments instead of or in addition to the treatment the patient was on.
Cohen suggests that PGx tests offer no clear clinical value over freely available and well-established protocols for drug selection with a reference to a number of recent documents.20,21 He also suggests that, should a clinician be unsure about drug choice, a psychiatry consultation costs substantially less money than PGx testing.
Additional comments of support
This contractor received numerous comments on this proposed policy. Of note, we received letters of support for coverage of PGx testing as described in this policy by prominent associations including the Association for Molecular Pathology, the College of American Pathologists, and the American Psychiatric Association.
Evidence for the utility of combinatorial PGx testing
A review of the available evidence found five combinatorial PGx tests for which outcome data has been published. These studies focus on testing for drug selection for psychiatric and neurological use, and include tests such as CNSDonse, Genecept, Neuropharmagen, Genesight, and NeuroIDGenetix.
A review of the published clinical studies performed demonstrated that, despite showing some promising findings, these studies had several shortcomings in that they failed to meet their primary endpoints,25-28 were not representative of the Medicare population,29-32 or had other significant shortcomings such as a possible biased sample population.26,29-31,33
Some of these studies yielded interesting observations, for example, Tanner et al.34 that demonstrated that greater improvement in treatment success was seen by primary care providers using a combinatorial PGx test than compared to psychiatrists who did not see a significant benefit, although there was no difference identified between the psychiatrist-treated and primary care treated groups. Many of these studies saw differences in performance from using these genetic tests vs. standard of care with the limitations noted above.
We found two meta-analyses of combinatorial PGx with overlap in the studies reviewed, the statistical techniques, and the results. Both meta-analyses included studies that are cited above. All the tests studied in these meta-analyses included CYP2D6 and CYP2C19. Both pooled data using a random effects model. The meta-analysis published by Rosenblatt, Lee, and McIntyre,35 pooled the results of 6 studies and found a pooled relative risk for treatment remission of 1.71 in favor of PGx testing, and a relative risk of 1.36 for remission, also favoring testing. A more recent meta-analysis by Bousman et al.36 included 5 studies and found a pooled relative risk for treatment remission of 1.74 in favor of PGx testing. A pooled relative risk for response was not reported. Both of these meta-analyses found a high level of heterogeneity, and the paper by Rosenblatt, Lee, and McIntyre suggested that this makes it difficult to assume that there is a class effect. Additionally, because combinatorial PGx tests rely on many of the same PGx biomarkers tested in single and multi-gene PGx tests, it has been postulated based on an observational study utilizing the GeneSight that there may be little gained from the proprietary algorithms in these tests above single and multi-gene panels.37 It should be restated here that our CAC experts were consistent in their lack of support for the use of combinatorial PGx and cited similar questions; additional comments received from societies such as the Association for Molecular Pathology, the College of American Pathologists, and the American Psychiatric Association similarly stated that the validity and utility of combinatorial testing has not been demonstrated. Of note, a more recent review submitted by the APA concludes that "there is insufficient evidence to support widespread use of combinatorial pharmacogenetic decision support tools at this point in time" and noted that "a high level of evidence has been achieved only for the cytochrome P450 genotype data".38