Psychiatrist, The Permanente Medical Group, Kaiser-Permanente San Francisco Medical Center, Clinical faculty, Department of Psychiatry, University of California, San Francisco.Dr. Hamilton has disclosed that he is an unpaid scientific advisor to 23andMe, which offers a commercial genome test. Dr. Balt has reviewed this interview and found no evidence of bias in this educational activity.
TCPR: Dr. Hamilton, why would a psychiatrist want or need to order genetic testing on a patient?
Dr. Hamilton: I tend to consider genetic testing for two specific types of cases. One is serial intolerability—poor tolerance of multiple medications, for instance, across SSRIs or even between classes of medications. This gets me wondering if a person has some perturbation in their metabolism; a genetic variant that might affect how they process these drugs. The other indication is when a patient shows resistance to antipsychotics or to antidepressants.
TCPR: Can you summarize briefly some of the genes that are tested in pharmacogenetic testing in psychiatry?
Dr. Hamilton: Because of their connection to a broad swath of psychotropic drugs, it really boils down to two cytochrome P450 genes: 2D6 and 2C19. This is because nearly all psychotropics are metabolized to some extent by the enzymes encoded by these genes. More importantly, there is ample genetic variation in these genes, specifically related to people being poor metabolizers, and these are fairly common in the general population. And with CYP2D6, there is an added level of complexity in that you can have individuals who are what we term ultra-metabolizers, meaning their gene products work at a much higher rate of efficiency than the normal level of functioning.
TCPR: You talk about people being normal metabolizers or poor metabolizers on these genes. What is the evidence that metabolizer status correlates with blood levels of medication and/or clinical outcome?
Dr. Hamilton: There has been a lot of effort put into looking at the correlation between drug levels and polymorphism status. The evidence is good in the case of tricyclic antidepressants. The classic study from the ‘90s showed that nortriptyline levels were highly correlated with metabolizer status for 2D6 (Dalén P et al, Clin Pharmacol Ther 1998;63(4):444–452). Those who were ultra-metabolizers had almost unmeasurable amounts of nortriptyline in their system. So there are a few clinical observations, such as this, but these are typically clinical pharmacokinetic laboratory studies. What’s really missing are large prospective studies looking at actual clinical outcomes based on metabolizer status.
TCPR: Do you think it is possible that there is a sort of intuitive pharmacogenetics going on all the time in psychiatry? That patients who are regular or poor metabolizers end up on the appropriate dose of medication through the normal process of monitoring and titrating medication?
Dr. Hamilton: Some research shows that, yes, clinicians frequently alter the dose of an individual’s medication in the direction that fits, using a sort of intuitive pharmacogenetics. An older study looked at the genotypes of people taking risperidone (Risperdal). Investigators looked at individuals’ doses of risperidone over time, and later looked at their genotypes. You would expect that individuals who were poor metabolizers probably couldn’t tolerate higher doses because it was cleared less efficiently. Sure enough, individuals who had poor-metabolizer status were placed on lower doses, even though the clinicians did not know anything about their P-450 enzymes (Mas S et al, 2012 Pharmacogenomics J;12(3):255–259).
TCPR: The massive STAR*D trial found no individual antidepressant strategy to be better than any other. Is it possible that applying pharmacogenetic strategies to patients in order to guide treatment might have led to better outcomes in a study like STAR*D?
Dr. Hamilton: My own work was involved with genetic studies of the STAR*D sample. We carried out a retrospective genome-wide association study (GWAS) of the STAR*D data. The results we found and published did not meet the standard levels of statistical significance, and were not robust enough to have warranted changing the treatment, even if we had that information before the STAR*D study began. There is an exception to that. We found that CYP2D6 or CYP2C19 metabolizer status did not predict response, and published that finding some time ago (Peters EJ et al, 2008 PLoS One, 3(4):e1872). However, in work that we did not publish then, we looked at 2D6 and 2C19 metabolizer status and found that the rates of intolerability—defined by STAR*D as whether an individual could continue taking the medication based on side effects—was correlated with 2D6 and 2C19 genotypes. Knowing that ahead of time may have been useful, because one of the primary predictors of antidepressant response in STAR*D was drug intolerability.
TCPR: Are there any guidelines for the use of pharmacogenetic testing to optimize drug therapy?
Dr. Hamilton: There are about 25 psychotropic drugs for which the FDA has guidance for pharmacogenetics (these can be found at http://1.usa.gov/1cCZvmr). It is an interesting list that I would urge clinicians to check out. Another resource is the Clinical Pharmacogenetics Implementation Consortium (CPIC), a partly NIH-funded group of researchers, in this field, who have come together to create guidelines (www.pharmgkb.org/page/cpic). They provide recommendations for a broad array of medications, including several tricyclic antidepressants.
TCPR: Isn’t it true that behavioral factors contribute to drug response in a way that might outweigh the genetic factors?
Dr. Hamilton: Absolutely. Genetics only make up a small part of the likelihood of drug response or tolerability, and other issues are strong contributors. Other exogenous substances, such as tobacco or dietary supplements, may interfere with the metabolism of antidepressants and other metabolic issues related to age, disease (eg, hepatic disorders), and gender can influence these things. Genetics likely plays some discrete, but measurable, role, but it needs to be taken into account with all these other nongenetic factors.
TCPR: You mentioned the genome-wide association study (GWAS) analysis. What is GWAS?
Dr. Hamilton: GWAS allows us to look not at just a few genes, but tens of thousands of genes and millions of genetic polymorphisms simultaneously. GWAS asks whether variations from numerous genes across the genome, each of which might contribute only a small amount individually, might together influence a phenotype. Also, the question can be asked whether combinations of variants from a multitude of genes influence medication response. In a recent article, we combined all the data from large European and American studies and it was still insufficient to see anything of strong statistical significance. The individual contribution to antidepressant response by variation from numerous genes is of such low amplitude that we can’t detect it with the numbers of samples that we have (GENDEP Investigators, MARS Investigators, STAR*D Investigators, Am J Psychiatry 2013;170(2):207–217). In the case of schizophrenia, a recent study suggested that there are over 100 variants that influence risk for this disorder. But it took tens of thousands of samples to find them. It is one thing to find someone with a particular disease phenotype, but to find someone who has gone through a clinical trial with a well-delineated treatment response phenotype—is much more difficult. So for our meta-analysis we only had several thousand individuals, which is probably insufficient to find the pattern of variation that would help in what you are describing for personalizing medication choice.
TCPR: Can you describe what pharmacogenetic tests are currently available to psychiatrists?
Dr. Hamilton: A great resource is a website called GeneTests.org. It is a compendium of genetic tests, not only for rare genetic disorders, but also more common tests as well. You can search for a genetic test and it will actually give you a list of all the different providers of this test. An alternative website, where practitioners can get information about genetic tests, their costs, and comparisons, is Nextgxdx.com. There are several commercial operations that sell tests for psychiatric applications. I am not recommending any of them, in any way, by mentioning them. They include Genomind, Pathway Genomics, Genelex, and Assurex Health.
TCPR: In addition to 2D6 and 2C19, there are other pharmacokinetic genes, as well as a number of pharmacodynamics genes, such as the serotonin transporter (SLC6A4), DRD2, and MTHFR, offered as a part of commercial tests. Do these also help to guide treatment options?
Dr. Hamilton: To put it simply, no. In my work studying all of these genes, as well as the work of many others, there have never been any consistent correlations between clinical response to medication and any of these particular genotypes. They are very appealing because they get at presumptive mechanisms for these drugs, but when you actually look at large samples for which there is statistical power to detect an association, the findings are either inconsistent or just don’t exist. The evidence just does not support any of them at this time.
TCPR: These tests have a reputation for being expensive. How much do they cost?
Dr. Hamilton: The prices for a relatively focused gene panel such as 2D6 and 2C19, in general, run in the hundreds of dollars. I am not aware of a strong push to get insurance companies to pay for these.
TCPR: Can you speak to the applications of commercial genome tests such as the one offered by 23andMe? Could a commercial test such as that give me the same information as these pharmacogenetic tests?
Dr. Hamilton: 23andMe offers an inexpensive genome test—about $99. I think it gives data on about a million genetic variants, including functional variants in 2C19. The information provided depends on the variant. For instance, the 2D6 region is so extremely complicated at the structural level that many of these commercial genome-wide tests such as 23andMe do not really test that gene very well, if at all. On the other hand, the information about 2C19 is in there, along with several other variants. Right now, the FDA does not allow 23andMe to give you any advice relating to the variant.
TCPR: Are you aware of any pharmaceutical developments capitalizing on specific pharmacodynamic polymorphisms?
Dr. Hamilton: There are examples from the cancer literature where specific drugs are targeted to particular mutations in tumors, but we are not there yet for psychotropics.
TCPR: Thank you, Dr. Hamilton.
PO Box 626, Newburyport MA 01950