Weighing the Benefits of Genetic Information in Clinical Psychiatry

Have you considered using pharmacogenomic testing in your practice?

Genetic information is slowly working its way into psychiatric practice and our patients’ daily lives. Pharmacogenomic panels can be ordered by psychiatrists, and genetic testing is being sold to the general public without any involvement of physicians or genetic counselors. The practicing psychiatrist needs to be informed about both of these types of services.

Genetic panels marketed to psychiatrists

When pharmacogenomic panels started to hit the market about 5 years ago, most general psychiatrists greeted this option with understandable skepticism. Now the use of these tests is growing, both in numbers of genes tested and in the number of psychiatrists utilizing them.

According to Daniel Dowd, PharmD, Vice President of Medical Affairs at Genomind, one of the two leading commercial pharmacogenomic services, close to 5000 clinicians are currently using Genomind’s panel, and the company has recently logged its 100,000th test. “What we’re actually seeing in practice is most psychiatrists are ordering this for the treatment-resistant patients with 2 or more treatment failures,” says Dr. Dowd. “But in the pediatric patient population, they’re more likely to order it for treatment-naive patients.”

John A. R. Grimaldi Jr, MD, Instructor at Harvard Medical School (HMS) and attending outpatient psychiatrist at Brigham and Women’s Hospital (BWH), says he has had many patients ask about these genetic panels, and 2 patients brought in genetic panels ordered by their previous psychiatrist. “There is a real need for guidance about the use of these ‘assays’,” says Dr. Grimaldi. “I can imagine that ‘genetic assays’ would be useful with treatment-resistant patients, if the assay were paid for by insurance and if there were an evidence base to support interpretation.”

As a new neuropsychiatry fellow, I started to use pharmacogenomic testing for some of my patients last year when I inherited and started to collect a patient caseload filled with very complicated people. Some had tried many different psychiatric medications with little effect, or with significant and sometimes surprising adverse effects.

My supervisor and mentor, Barry Fogel, MD, Professor of Psychiatry at HMS and staff neurologist and psychiatrist at BWH, had asked me if I had considered pharmacogenomic testing. I objected that I did not know anyone who had ever used this testing, and that I did not know how to order it or how much this would cost my patients. Dr. Fogel gently pointed out that these were not insurmountable barriers.

Options for testing

When I looked into the options for pharmacogenomic testing, I found multiple companies offering similar panels, some of which had booths at the American Psychiatric Association meeting in May 2017. The biggest and most well-established providers of pharmacogenomic testing are GeneSight and Genomind.

GeneSight currently offers 4 panels: psychotropic, analgesic, ADHD, and methylene tetrahydrofolate reductase (MTHFR). The GeneSight website states that its “Psychotropic laboratory developed test analyzes how patients’ genes may affect his/her metabolism and response to FDA-approved medicines . . . .”

The psychotropic panel looks at genes for liver enzymes that are responsible for the metabolism of most neuropsychiatric medications. It tests for various gene alleles and classifies the possible phenotypes as ultra-rapid, extensive, intermediate, or poor metabolizers. The panel also tests for a few genes not related to hepatic metabolism of psychotropic drugs, which are receptor polymorphisms with potential pharmacodynamic implications.

Genomind has one panel, the “Genecept Assay.” The assay includes genes encoding the serotonin transporter, calcium channel, sodium channel, serotonin receptor 2C, melanocortin 4 receptor, dopamine 2 receptor, catechol-O-methyltransferase, alpha-2A adrenergic receptor, MTHFR, brain-derived neurotrophic factor, μ-opioid receptor, and glutamate receptor, as well as a number of the hepatic cytochromes. According to Dr. Dowd, Genomind has received the most positive feedback from clinicians regarding the serotonin transporter gene (SLC6A4) and the 2 weight-gain–related genes on the panel (5HT2C and MC4R). Dr. Dowd asserts that certain serotonin transporter gene alleles identify people with lower response and remission rates and more adverse effects when treated for depression with SSRIs and that certain alleles of 5HT2C and MC4R imply a significantly higher risk of weight gain if the patient is treated with an atypical antipsychotic drug.

Words of caution

Shaun Purcell, PhD, Associate Professor of Psychiatry at HMS and a genetic epidemiologist who was involved in some of the studies to identify genes as risk factors for mental disorders, believes psychiatrists have ample cause to be cautious about using genetic testing. “To me, the evidence is underwhelming to say the least. It’s definitely concerning that these kinds of tests are being pushed prematurely.”

For the non-metabolic genes on the panel, Dr. Purcell says, “I’d question the basic assumptions and science behind the panel. As far as I’m aware, these genes are coming from studies of disease risk.” He adds, “It’s a different thing to have an association in the literature and to have it be clinically meaningful.”

Dr. Purcell continues: “The associations of major psychiatric illness with common variants in CACNA1C are unambiguously sound. They say nothing about what implications these variants have, if any, for guiding treatment.” He adds, “There is a fundamental disconnect between what is being claimed and the extant evidence base for CACNA1C.” He also questions the premise that any of the other non-metabolic genes tested on the GeneSight and Genomind panels have clinical utility. “There’s no evidence that these variants are going to be clinically actionable in terms of guiding treatment.”

Dr. Fogel, who is generally in favor of clinicians considering pharmacogenomic testing for difficult-to-treat patients, agrees with Dr. Purcell that the clinical relevance of some of the genes on the panels is questionable. He says, “Will getting a gene associated with BDNF actually make more of a difference than getting a BDNF level? I don’t think so.” Regarding Genomind’s listed “treatment impacts” for the MTHFR and BDNF genes, Dr. Fogel says, “A methylfolate supplement and exercise would usually be good for a patient no matter what their genotype.”

Of course, the evidence is different for each gene on the panels. For example, take the serotonin transporter gene, SLC6A4, which is heavily studied. The short form of the serotonin transporter-linked polymorphic region is associated with reduced activity of the serotonin transporter (low expression) and with higher levels of intrasynaptic serotonin. The long form (high expression) is associated with increased activity of the serotonin transporter and lower levels of intrasynaptic serotonin.

This variation has been associated with affective disorders and with pathological behaviors and personality traits related to anxiety, impulsivity, and stress.1 For example, those with a low-expression form of the 5-HTTLPR serotonin transporter gene have a bias to pay attention to both negative and positive affective pictures.2 This suggests that the serotonin transporter gene might be a plasticity gene rather than a vulnerability gene-the low-expressing form allowing people to tune in to the affective significance of their surroundings.

Relevance to prescribing

More relevant to pharmacogenetic testing is the question of whether variations in the serotonin transporter gene can guide prescribing. The literature on this is fairly robust, with 33 studies included in the most recent meta-analysis in 2012. Of 3 meta-analyses published between 2007 and 2012, 2 found that the long/long (l/l) genotype in Caucasians was associated with response to antidepressants, mostly SSRIs, and all 3 found that carrying the l allele was associated with remission.3 Whether the l/l genotype predicts antidepressant response in Asians is less clear.

A 2016 pilot study of 35 patients showed that the l allele was associated with better response to desvenlafaxine.4 The genetic connection to treatment response may not only differ by ethnicity, but by age. A 2015 study of 234 patients with late-life depression showed that at least one short allele was associated with about a 2-fold probability of having a good SSRI response.5 Although the serotonin transporter gene has literature supporting its clinical relevance, there does not seem to be a clear, simple connection between genotype and treatment response.

But what about the CYP liver enzyme gene variations? The relationship of these liver enzymes to drug metabolism is not in question. “We cannot say who is going to respond to what,” Dr. Fogel says. “But the drug metabolism is pretty well established, and there are genotypes that imply someone is going to be a slower or faster metabolizer of certain drugs. That may be relevant in clinical situations where getting to a therapeutic drug level or avoiding a drug interaction or an unintentional overdose is really going to be a problem for the patient.” He adds, “You have some very specific situations where you can say it really makes sense to have advance intelligence on what sort of metabolizer this person with this drug will be.”

For my patients, getting information about drug metabolism from the genetic panels helped me build a therapeutic alliance and/or choose which drug to try next when the first drug I prescribed was not effective. One patient’s results showed slow metabolism for every psychotropic she had ever tried. She found this validating, as she had experienced an unusual amount of adverse effects, and we were able to pick a new option from the short list of medications she metabolizes well. Another patient felt reassured when his panel indicated that his metabolism was normal for almost everything. Thereafter, he expressed fewer concerns about trying new medications. A third patient felt validated that her testing supported her intolerance of and lack of response to SSRIs, and she was more willing to try other classes of antidepressants.

John F. Sullivan, MD, Instructor at HMS and neuropsychiatrist at BWH, has ordered GeneSight panels for 3 of his patients who “would have dramatic adverse reactions to tiny doses of anything, regardless of the medication’s mechanism. One of these patients had panic attacks to his multivitamin.” Dr. Sullivan found the results of the genetic testing useful to demonstrate that the severe adverse reactions his patients had to a multitude of medications were not related to a biologic process.

Laura Safar, MD, also an Instructor at HMS and neuropsychiatrist at BWH, has ordered panels for several patients and found that the information helped her understand patients’ response to medication adverse effects. The panels either validated patient concerns or, in one of her cases, supported her impression that the patient’s high levels of adverse effects to very low doses of medications were due to anxiety

Genetic panels and patient outcomes

Do genetic panels change patient outcomes? The largest GeneSight study was conducted in 2013, when the company had only a 5-gene pharmacogenomics test. This open-label study compared 114 patients whose clinicians had access to the GeneSight report (“guided”) with 113 patients whose clinician could not see the report (“unguided”). After 8 weeks, the guided group had a 70% greater improvement in depression scores, as well as better remission rates. Participants in the unguided group who at baseline were prescribed a medication that was “most discordant with their genotype” experienced the least improvement.6

Genomind did an unblinded study of its 10-gene panel in 2013, with 685 patients and a variety of diagnoses. The study showed that 91% of patients with 2 or more prior treatment failures had clinically measurable improvement, but the study was limited by having no treatment-as-usual comparison group.7

It is fair to say that there is not a large body of evidence to prove that ordering genetic panels affects treatment decisions or the course of disease.

Jane Erb, MD, Assistant Professor at HMS and Clinical Director of the Brigham Depression Center, has not incorporated genetic testing into her practice because “as of a couple of years ago, the data were not convincing that the information improved outcomes.” She says, “I have seen a handful or two of patients who have had testing run by others, and I’ve reviewed the results, which did not seem to explain the patients’ medication responses.”

Gaston Baslet, MD, Assistant Professor at HMS and attending neuropsychiatrist at BWH, says genetic testing “needs to be put in context with many other clinical factors that are usually going to be the ones directing my decision, such as factors obtained from a detailed med history.” He adds, “I therefore don’t think that the amount of time and effort in getting these results would be contributing enough to my decision-making to justify their regular use. Of course, there are exceptions to every rule.”

Putting the genetic information into clinical context is very important. A study published in January 2017 showed that less-experienced psychiatrists tended to be more influenced by genetic information than older psychiatrists.8 The study presented 67 psychiatrists with patients’ pre- or post-treatment scores on the Positive and Negative Syndrome Scale for 2 hypothetical treatments for schizophrenia. Psychiatrists were also informed whether the patient possessed a genotype linked to hyper-responsiveness to one of the treatments, and they were asked to recommend one of these two treatments. Less-experienced psychiatrists were inclined to believe the genetic information more than the clinical report of how patients responded to different drugs in the past, and to recommend treatment based on the genotype. McMichael and colleagues concluded that “psychiatrists and other clinicians should be cautious about allowing a patient’s genetic information to carry unnecessary weight in their clinical decision-making.”

Barriers to adoption of pharmacogenomics include changes in workflow and the time and effort involved in signing up at the company’s website, ordering test kits, obtaining cheek swabs, and sending in the samples. However, for both GeneSight and Genomind panels, collecting the samples was easy-a quick cheek swab in the office-and results were back within 2 weeks. Cost is also a major concern. Both companies do their best to get insurance companies to pay and promise that the most a patient will ever be charged (even with no insurance coverage) is in the low $300s.

In the future, we should expect expansion of the use of pharmacogenomic testing. The industry is growing rapidly, and in the next year or two the panels are expected to have more variants. There may be information on the genetic panels that can help determine how likely a patient is to respond to ECT or transcranial magnetic stimulation, but so far these are relatively small studies and need to be replicated.

Direct-to-consumer genetic information

With GeneSight and Genomind, the business model is for physicians to order the panels. But, there are services that offer direct-to-consumer genetic testing, such as 23andMe.

Last year, 23andMe was like fast food: fairly quick, kind of fun, and with little substance. It provided carrier status for 39 genetic diseases and not much else in the way of health information.

Before 2013, 23andMe used to provide a lot of clinical data, from carrier status for hereditary cancers to information about drug metabolism. Dr. Purcell comments that the original 23andMe genetic testing “included some patently bogus reports for psychiatric diseases including schizophrenia.”

In 2013, the FDA cracked down on the company for not getting proper approval to market what the FDA considered to be a medical device. 23andMe responded by significantly slimming down their panel. Now, they are building it back up. This year, 23andMe has a few new “Genetic Health Risk Reports”: the LRRK2 and GBA genes for Parkinson disease, the F5 and F2 genes for hereditary thrombophilia, and the SERPINA1 gene for alpha-1 antitrypsin deficiency. As a physician who treats dementia, I was surprised and dismayed to see that the APOE Alzheimer risk gene has returned to 23andMe after being removed in 2013. In general, though, health information is limited.

To get an idea of what these genetic tests provide, I decided to try 23andMe. What did I learn? I have fewer Neanderthal variants than 88% of customers. I am genetically more likely to have a “photic sneeze,” or sneeze in the sunlight (and I do). I am less likely to be a deep sleeper and am more likely to move in my sleep.

I was not surprised to find out that I am 63.2% British and Irish; a slow coffee metabolizer; unlikely to have a cleft chin, unibrow, dimples, or a widow’s peak (indeed, I don’t); and that my earlobes are likely to be detached (they are). I was very surprised to learn that I am more likely than others to be able to smell the asparagus metabolite in urine, though I can say with complete honesty that my curiosity about myself had never once strayed toward that question.

There is a way around the limitations of 23andMe: download your raw data, then upload it to another program to tease out the health information. The program I used is Promethease, which costs $5 to $10 depending on the source of the DNA. Promethease describes itself as “a literature retrieval system that builds a personal DNA report based on connecting a file of DNA genotypes to the scientific findings cited in SNPedia.”

The Promethease report is a far cry from the simplistic, jargon-free, and sparsely populated pages of the 23andMe website. For each gene variant, Promethease provides information about the frequency of that genotype in your population (eg, Caucasian), the location of the gene, and the number of papers published.

As with 23andMe, the information in Promethease was often amusing. For example, I have an allele that was associated with good memory in a Swiss cohort. I unfortunately do not have a gene associated with longevity in Japanese men. Some of the “genos” were helpful or enlightening: my single copy of the C677T allele of MTHFR puts me at some risk of having low vitamin B12 and folic acid. Of course, some were anxiety-provoking, like risk factors for cardiovascular disease, diabetes, and various inflammatory and neurological diseases.

Although I am of the general persuasion that knowledge is power, I find it reassuring that, even as a doctor with some (albeit limited) knowledge of genetics, it was not a simple process for me to find my own genetic risk factors. While some patients could certainly wade through their own Promethease data and glean appropriate or useful information, most people will likely not go beyond the information available on the 23andMe website.


Dr. Allen is a Clinical Fellow in Neuropsychiatry, Brigham and Women’s Hospital, Boston, MA. Dr. Allen reports no conflicts of interest concerning the subject matter of this article.


1. Serretti A, Calati R, Mandelli L, et al. Serotonin transporter gene variants and behavior: a comprehensive review. Curr Drug Targets. 2006;7:1659-1669.

2. Fox E, Zougkou K, Ridgewell A, Garner K. The serotonin transporter gene alters sensitivity to attention bias modification: evidence for a plasticity gene. Biol Psychiatry. 2011;70:1049-1054.

3. Porcelli S, Fabbri C, Serretti A. Meta-analysis of serotonin transporter gene promoter polymorphism (5-HTTLPR) association with antidepressant efficacy. Eur Neuropsychopharmacol. 2012;22:239-258.

4. Ng CH, Bousman C, Smith DJ, et al. A prospective study of serotonin and norepinephrine transporter genes and the response to desvenlafaxine over 8 weeks in major depressive disorder. Pharmacopsychiatry. 2016;49:210-212.

5. Seripa D, Pilotto A, Paroni G, et al. Role of the serotonin transporter gene locus in the response to SSRI treatment of major depressive disorder in late life. J Psychopharmacol. 2015;29:623-633.

6. Hall-Flavin DK, Winner JG, Allen JD, et al. Utility of integrated pharmacogenomic testing to support the treatment of major depressive disorder in a psychiatric outpatient setting. Pharmacogen Genom. 2013;23:535-548.

7. Brennan FX, Gardner KR, Lombard J, et al. A naturalistic study of the effectiveness of pharmacogenetic testing to guide treatment in psychiatric patients with mood and anxiety disorders. Prim Care Companion CNS Disord. 2015;17(2). doi: 10.4088/PCC.14m01717.

8. McMichael AJ, Boeri M, Rolison JJ, et al. The influence of genotype information on psychiatrists’ treatment recommendations: more experienced clinicians know better what to ignore. Value Health. 2017;20:126-131.