Do genetic tests for psychiatric medication work?
a cheek swab that tells your prescriber which antidepressant will work is one of the most appealing ideas in psychiatry. more than half a million of these tests were run by one company last year. this page lays out what is sold, what the randomised trials found, what the regulators and guideline bodies say, and the one genetic test in psychiatry that really does matter.
the short version
- the tests read two kinds of gene. liver enzyme genes change how fast you clear a drug, and that part is real. the other genes are sold as predicting whether a drug will work, and the guideline body says the evidence does not support that.[9]
- five randomised trials of commercial panels are below. four missed their primary goal. the fifth found a small gain that faded.
- pooled together the trials lean in favor of testing, and the reviewers rate the certainty very low, because the prescribers always knew who had been tested.[6]
- one genetic test in psychiatry has everyone’s backing: an immune gene check before carbamazepine. nobody runs ads for it.[20]
what is being sold
Myriad Genetics told investors it ran 537,000 GeneSight tests in 2025, up from 507,000, for $144.1 million, 17% of the company’s revenue.[14] the company’s consumer site says 98% of patients pay $330 or less.[24] other companies sell larger panels, some of them straight to the public.
| test | what it reads | who orders | price, as the company states it |
|---|---|---|---|
| GeneSight (Myriad) | 14 genes plus one reported for information only; a proprietary weighted algorithm; green, yellow and red bins | clinician only | “98% of patients pay $330 or less”; $0 typical on Medicare and Medicaid |
| Genomind | 26 genes, including MTHFR, COMT, BDNF, ANK3 and CACNA1C | clinician, and an at-home product | $599 self-pay for the at-home test |
| ClarityX Mindwell | 39 genes, “120+ mental health medications” | direct to consumer, with a physician reviewing the order | $399 list |
| OneOme RightMed | 27 genes, more than 100 medications | clinician only | not published |
| 23andMe pharmacogenetic reports | 33 variants, FDA-authorised in 2018 | consumer kit | FDA: not to be used to start, stop or change any treatment, and results must be confirmed by another test |
prices and panels read from each company’s own site on 2026-09-20. the 23andMe limits are the FDA’s.[25] more genes on a panel is not more evidence: most of the added genes have no prescribing guideline behind them.
what the randomised trials found
a trial names its main goal before it starts. that is the primary outcome, and it is the fairest single test of whether something worked. here is each trial’s primary outcome, in its own numbers.
GUIDED · GeneSight · 1,167 patients[1]
primary goal not metprimary outcome: symptom improvement at 8 weeks: 27.2% vs 24.4%, p = 0.107
response (26.0% vs 19.9%) and remission (15.3% vs 10.1%) were significant secondary results. several authors were company employees.
PRIME Care · commercial panel, VA · 1,944 patients[2]
primary goal met, small and fadedprimary outcome: remission over 24 weeks: odds ratio 1.28, about 2.8 points. at week 24: 1.5 points, p = .45
the authors’ words: “small nonpersistent effects.” fewer drugs with predicted interactions were prescribed. funded by the VA.
Perlis 2020 · Genomind · 304 patients[3]
primary goal not metprimary outcome: depression score change at 8 weeks: p = .53
the funding field of the record reads Genomind, Inc.
GAPP-MDD · GeneSight[4]
primary goal not metprimary outcome: improvement, response and remission all favored testing, none significant
the authors say the trial was underpowered. company authors.
Pérez 2017 · Neuropharmagen · 316 patients[5]
primary goal not metprimary outcome: sustained response over 12 weeks: 38.5% vs 34.4%, p = 0.47
a secondary responder rate at week 12 was significant.
the case for the tests
it is not nothing. when reviewers pool the trials, tested patients reach response and remission somewhat more often. a 2025 pooling of thirteen trials put response about 23% to 29% higher in relative terms, though remission was no longer significant at twelve weeks.[7] a 2026 analysis split the trials by product: the three GeneSight trials pooled to a significant result, and all the other panels pooled to a result that was not.[8] the American Psychiatric Association’s task force, while finding the data insufficient for widespread use, said the tests may be informative in some situations, particularly in predicting side effects.[10]
the problem that runs through all of it
in these trials the patient and the rater did not know who had been tested. the prescriber did, because the test is a coloured report handed to the person choosing the drug. a prescriber who believes in the report may follow up sooner, adjust faster and sound more confident. one rapid review rated the certainty of the pooled benefit very low for that kind of reason.[6] the 2025 pooling also found that the effect stopped growing as newer, bigger panels were added: “broader panels offering minimal incremental gains.”[7]
what the guideline bodies and the FDA say
- the guideline body for gene and drug pairs (CPIC) gives dosing advice for the liver enzyme genes CYP2D6, CYP2C19 and CYP2B6 with SSRIs. for the serotonin transporter and receptor genes that most panels also report, SLC6A4 and HTR2A, it says the data “do not support their clinical use in antidepressant prescribing.”[9]
- the International Society of Psychiatric Genetics supports testing the two liver enzyme genes and the two immune genes, and says it does not recommend direct-to-consumer genetic testing for medical purposes in people with psychiatric illness.[11]
- the American Psychiatric Association’s task force concluded in 2018 that there were “insufficient data to support the widespread use of combinatorial pharmacogenetic testing in clinical practice.”[10]
- the FDA keeps a public table of gene and drug pairs in three tiers. citalopram, venlafaxine, vortioxetine, aripiprazole and atomoxetine sit in the top tier, where the data support a change in management. escitalopram, paroxetine, fluvoxamine, amitriptyline, nortriptyline and risperidone sit in the bottom tier: an effect on drug levels only.[12] in 2019 the agency warned one laboratory that “the relationship between CYP2C19 genotype and drug response to escitalopram and sertraline is not established.”[13]
- Myriad’s own annual report says the FDA requested changes to the GeneSight offering in 2019, that the company disagreed, submitted a proposal, and has not implemented it.[14] these tests are laboratory-developed tests, a category the FDA does not review before sale. a 2024 rule that would have changed that was struck down in court in 2025.
five things the brochure leaves out
1. the genes predict drug level, not whether the drug works
researchers pooled individual data on 5,843 people with depression across 13 studies. how fast a person’s CYP2C19 or CYP2D6 enzymes ran did not predict how much their depression improved. one signal for remission did not survive correction for the number of tests run.[18]
2. your other medicines can overrule your genes
several common psychiatric drugs block the very enzymes the test reads. in 900 patients taking venlafaxine, 4% were poor metabolisers by gene and 27% were poor metabolisers in practice.[16] in a study of 15,000 psychiatric patients, 42% of those with medication records had this kind of shift.[17] a report that reads only DNA cannot see it.
3. different companies disagree about the same person
two researchers sent five patients’ DNA through four commercial tools. the tools agreed on antidepressant recommendations 56% of the time. among recommendations flagged as actionable by more than one tool, 19% conflicted. their conclusion: the tests “cannot be assumed to be equivalent or interchangeable.”[15]
4. the large gene hunts have come up empty
the biggest genome-wide study of antidepressant response found no single variant that reached significance. common variants together explained about 13% of the difference in remission, with a wide margin of error, and none of the difference in percentage improvement.[19]
5. the business took a hit when an insurer looked closely
in 2024 UnitedHealthcare stopped covering multi-gene panels including GeneSight under its commercial plans. Myriad reported a 20% drop in revenue per test and wrote down $91.2 million of goodwill and $71.8 million of intangible assets on the unit in 2025.[14] test volume still rose 6%.
the genetic test that does matter
before carbamazepine or oxcarbazepine, a check for two immune gene variants, HLA-B*15:02 and HLA-A*31:01, is recommended by the guideline body. people who carry them are at much higher risk of Stevens-Johnson syndrome and toxic epidermal necrolysis, rare skin reactions that can kill.[20] it is a narrow test for a specific danger, and it is the clearest win genetics has given psychiatry so far. if you are about to start one of those two drugs, it is a fair thing to ask about.
other “biological” tests you may be offered
- MTHFR testing, often paired with L-methylfolate. the medical genetics college says the test “has minimal clinical utility and, therefore should not be ordered.”[22] of the two L-methylfolate trials, the first (148 people) found no difference and the second (75 people, higher dose) was positive.[21]
- brainwave (EEG) tests for ADHD. the American Academy of Neurology advises that the theta to beta ratio “should not replace a standard clinical evaluation” and warns of harm from an “unacceptably high false-positive diagnostic rate.”[23]
- urine neurotransmitter, cortisol and gut microbiome tests. we found no trial showing that any of them improves a treatment choice. that is an absence of evidence, which is different from proof they fail.
questions worth asking before you pay
- which genes on this report have a dosing guideline behind them, and which do not?
- do any of my current medicines block CYP2D6 or CYP2C19? would that change what this report says?
- what would you do differently with a red, a yellow or a green?
- what will it cost me if my insurer declines?
a test result is one input. it is never a reason to stop or switch a medication without your prescriber. for how drug companies and test makers reach prescribers in the first place, see how psychiatric drug marketing feeds itself.
questions people ask
Does the GeneSight test work?
Its main randomised trial, GUIDED, did not meet its primary goal: symptom improvement at eight weeks was 27.2% with the test and 24.4% without, a difference that was not statistically significant. Two secondary measures, response and remission, did favor the test. The prescribers in the trial knew which patients had been tested, which makes secondary results hard to read.
What did the largest trial find?
PRIME Care followed 1,944 veterans. Tested patients were less often prescribed a drug with a predicted gene interaction, and remission over 24 weeks was slightly more likely, about 2.8 percentage points. At week 24 itself there was no difference. The authors called the effect on remission small and nonpersistent.
Is any genetic test in psychiatry clearly worth doing?
Yes. Testing for HLA-B*15:02 and HLA-A*31:01 before carbamazepine or oxcarbazepine is recommended by the guideline body, because those variants are tied to rare, life-threatening skin reactions. Liver enzyme genes, CYP2D6 and CYP2C19, also have dosing guidelines for some drugs.
Should I take an MTHFR test for depression?
The American College of Medical Genetics and Genomics says MTHFR polymorphism testing has minimal clinical utility and should not be ordered. In the two L-methylfolate trials that are cited for it, one was negative and one was positive, and patients were not selected by MTHFR gene.
I already have a report. What do I do with it?
Bring it to your prescriber and ask which lines rest on the liver enzyme genes and which rest on the other genes. Ask whether any of your current medicines block those enzymes, because that can matter more than the gene. Do not stop or switch a medication on the strength of a colour on a report.
what we could not verify, and left out
the wording of Medicare’s coverage policy, any position from the UK’s NICE, how much is spent advertising these tests, and the price Myriad paid for the company that built GeneSight. if you can point us to a primary source for any of them, we will add it.
sources
- Greden JF, et al. Impact of pharmacogenomics on clinical outcomes in major depressive disorder in the GUIDED trial. J Psychiatr Res. 2019;111:59-67. PMID 30677646. https://pubmed.ncbi.nlm.nih.gov/30677646/
- Oslin DW, et al. Effect of pharmacogenomic testing for drug-gene interactions on medication selection and remission of symptoms in major depressive disorder: the PRIME Care randomized clinical trial. JAMA. 2022;328(2):151-161. PMID 35819423. https://pubmed.ncbi.nlm.nih.gov/35819423/
- Perlis RH, et al. Randomized, controlled, participant- and rater-blind trial of pharmacogenomic test-guided treatment versus treatment as usual for major depressive disorder. Depress Anxiety. 2020;37(9):834-841. PMID 32383277. https://pubmed.ncbi.nlm.nih.gov/32383277/
- Tiwari AK, et al. Clinical utility of combinatorial pharmacogenomic testing in depression: a Canadian patient- and rater-blinded, randomized, controlled trial (GAPP-MDD). Transl Psychiatry. 2022;12:101. PMID 35288545. https://pubmed.ncbi.nlm.nih.gov/35288545/
- Pérez V, et al. Efficacy of prospective pharmacogenetic testing in the treatment of major depressive disorder. BMC Psychiatry. 2017;17:250. PMID 28705252. https://pubmed.ncbi.nlm.nih.gov/28705252/
- Bunka M, et al. Evaluating treatment outcomes in pharmacogenomic-guided care for major depression: a rapid review and meta-analysis. Psychiatry Res. 2023;321:115102. PMID 36780865. https://pubmed.ncbi.nlm.nih.gov/36780865/
- Zhang Y, et al. Comparative effectiveness of pharmacogenomic-guided versus unguided antidepressant treatment in major depressive disorder: new insights from subgroup and cumulative meta-analyses. BMJ Ment Health. 2025;28(1):e301726. PMID 40854804. https://pubmed.ncbi.nlm.nih.gov/40854804/
- Pullano I, et al. Clinical effectiveness of pharmacogenomic-guided antidepressant treatment in adult major depressive disorder: systematic review and exploratory platform-aware meta-analysis. J Psychiatr Res. 2026;201:549-562. PMID 42462453. https://pubmed.ncbi.nlm.nih.gov/42462453/
- Bousman CA, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2D6, CYP2C19, CYP2B6, SLC6A4, and HTR2A genotypes and serotonin reuptake inhibitor antidepressants. Clin Pharmacol Ther. 2023;114(1):51-68. PMID 37032427. https://pubmed.ncbi.nlm.nih.gov/37032427/
- Zeier Z, et al. Clinical implementation of pharmacogenetic decision support tools for antidepressant drug prescribing (American Psychiatric Association task force). Am J Psychiatry. 2018;175(9):873-886. PMID 29690793. https://pubmed.ncbi.nlm.nih.gov/29690793/
- International Society of Psychiatric Genetics. Genetic Testing Statement (approved March 11, 2019). https://ispg.net/genetic-testing-statement/
- U.S. Food and Drug Administration. Table of Pharmacogenetic Associations (updated September 10, 2026). https://www.fda.gov/medical-devices/precision-medicine/table-pharmacogenetic-associations
- U.S. Food and Drug Administration. Warning letter to Inova Genomics Laboratory, April 4, 2019. https://www.fda.gov/inspections-compliance-enforcement-and-criminal-investigations/warning-letters/inova-genomics-laboratory-577422-04042019
- Myriad Genetics, Inc. Annual report on Form 10-K for the year ended December 31, 2025 (filed February 24, 2026). U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/899923/000089992326000018/mygn-20251231.htm
- Bousman CA, Dunlop BW. Genotype, phenotype, and medication recommendation agreement among commercial pharmacogenetic-based decision support tools. Pharmacogenomics J. 2018;18(5):613-622. PMID 29795409. https://pubmed.ncbi.nlm.nih.gov/29795409/
- Preskorn SH, et al. Cytochrome P450 2D6 phenoconversion is common in patients being treated for depression. J Clin Psychiatry. 2013;74(6):614-621. PMID 23541126. https://pubmed.ncbi.nlm.nih.gov/23541126/
- Patel JN, et al. Pharmacogenomic insights in psychiatric care: uncovering novel actionability, allele-specific CYP2D6 copy number variation, and phenoconversion in 15,000 patients. Mol Psychiatry. 2024;29(11):3495-3502. PMID 38783055. https://pubmed.ncbi.nlm.nih.gov/38783055/
- Li D, et al. Metabolic activity of CYP2C19 and CYP2D6 on antidepressant response from 13 clinical studies using genotype imputation. Transl Psychiatry. 2024;14:296. PMID 39025838. https://pubmed.ncbi.nlm.nih.gov/39025838/
- Pain O, et al. Identifying the common genetic basis of antidepressant response. Biol Psychiatry Glob Open Sci. 2022;2(2):115-126. PMID 35712048. https://pubmed.ncbi.nlm.nih.gov/35712048/
- Phillips EJ, et al. CPIC guideline for HLA genotype and use of carbamazepine and oxcarbazepine: 2017 update. Clin Pharmacol Ther. 2018;103(4):574-581. PMID 29392710. https://pubmed.ncbi.nlm.nih.gov/29392710/
- Papakostas GI, et al. L-methylfolate as adjunctive therapy for SSRI-resistant major depression: results of two randomized, double-blind, parallel-sequential trials. Am J Psychiatry. 2012;169(12):1267-1274. PMID 23212058. https://pubmed.ncbi.nlm.nih.gov/23212058/
- Hickey SE, et al. ACMG practice guideline: lack of evidence for MTHFR polymorphism testing. Genet Med. 2013;15(2):153-156. PMID 23288205. https://pubmed.ncbi.nlm.nih.gov/23288205/
- Gloss D, et al. Practice advisory: the utility of EEG theta/beta power ratio in ADHD diagnosis (American Academy of Neurology). Neurology. 2016;87(22):2375-2379. PMID 27760867. https://pubmed.ncbi.nlm.nih.gov/27760867/
- Myriad Genetics. GeneSight test cost (company page). https://genesight.com/cost/
- U.S. Food and Drug Administration. Direct-to-consumer tests (23andMe pharmacogenetic reports, DEN180028, and their limits). https://www.fda.gov/medical-devices/in-vitro-diagnostics/direct-consumer-tests
adam