Genetic Testing in Psychiatry · Section A · Chapter 1 · Should I Test?
Should I Order a Pharmacogenomic Panel?
A pharmacogenomic panel can tell you how a patient is likely to metabolize specific drugs. It does not reliably tell you which antidepressant will work. This chapter covers what the randomized trials showed, what guideline bodies and Medicare say about ordering, which patients a result can help, and when to skip the test.
The 30-second version
- Panels do not reliably identify which antidepressant will work. In depression trials, test-guided care changed which drugs were prescribed more reliably than it changed outcomes, and benefits varied by outcome and time point. GUIDED, sponsored by the test's manufacturer, missed its primary endpoint. PRIME Care found a small improvement in remission across 24 weeks but no significant difference at week 24. ADOPT PGx did not meet its primary endpoint at 3 months but found higher remission at 6 months, a secondary outcome. In the APA work group's 2024 review, 3 of 11 newer trials met their primary outcome, and all had significant limitations.
- What a result can do. For specific drug–gene–phenotype combinations, guidelines tell you how to adjust the dose or choose another drug. CPIC, for example, covers CYP2C19 with citalopram, escitalopram and sertraline, CYP2B6 with sertraline, and CYP2D6 with paroxetine, venlafaxine and vortioxetine. CPIC, DPWG and FDA labeling can disagree: for venlafaxine, CPIC gives an optional recommendation to consider an alternative in CYP2D6 poor metabolizers, while DPWG advises avoiding it in poor and intermediate metabolizers. Name the gene, phenotype, drug and source rather than combining recommendations. CPIC states that its guidelines assume the result is already in hand and do not discuss whether to order the test.
- There is no validated number of failed trials. No trial established how many failed antidepressants justify testing. Consider testing when a result could change a specific prescribing decision you face now or soon, after weighing existing results, alternatives and cost. A drug–gene recommendation makes testing an option, not a requirement; DPWG, for example, advises considering testing for individual patients rather than all patients. Medicare contractors apply the same logic: one policy does not cover testing on the basis of a diagnosis alone.
- Genetic safety checks are a different question. The carbamazepine label calls for HLA-B*15:02 screening before treatment in patients with ancestry in at-risk populations, and separately warns about HLA-A*31:01. Valproate is contraindicated in known POLG-related mitochondrial disorders and in children under 2 suspected of having one. These targeted safety checks are separate from broad antidepressant-selection testing. Chapter 2 covers them.
- Don't order a test to answer a question it can't answer. Do not use SLC6A4, HTR2A, ANK3 or MTHFR results alone to select or dose psychotropic medication. CPIC found the evidence for SLC6A4 and HTR2A mixed and/or insufficient to support clinical use for the serotonin reuptake inhibitor antidepressants it reviewed (CPIC level C, no recommendation) [13]; MTHFR is covered in Chapter 8; SLC6A4, HTR2A and ANK3 in Chapter 9. In children and adolescents, AACAP advises clinicians to avoid using pharmacogenetic testing to select psychotropic medications.
At the visit: deciding whether to order a test
About 2 min to read · the detailed chapter follows below
Do now
- Name the drug you are using or about to prescribe, and the decision the result would change.
- Check whether that drug has a CPIC, DPWG or FDA label recommendation for a gene the test covers (Chapters 5 to 7).
- Ask about side effects at low doses and about no response at an adequate dose and duration.
- Check whether the patient has already been tested, and whether that report covers what you need (Chapter 3).
- Check coverage and the patient's likely out-of-pocket cost before ordering.
Do not
- Promise that the test will find the right medicine.
- Change an effective, tolerated drug solely because of a report’s color category. Review any actionable safety finding first.
- Use SLC6A4, HTR2A, ANK3 or MTHFR results alone to select or dose psychotropic medication.
- Treat a report's color-coded category as a prescribing guideline.
- Delay needed psychiatric care while waiting for an optional panel. Where a genetic safety check is indicated, use an alternative drug until it is done.
Five questions to ask
- What decision will the result change? If you cannot name one, the test is unlikely to help.
- Which drug am I using or about to prescribe, and does it have a gene–drug recommendation? Check the drug against CPIC, DPWG and the FDA label, not against the report's categories.
- Has the patient had side effects at low doses, or no response at adequate doses? Either pattern can reflect altered metabolism, but also nonadherence, too short a trial or an interacting drug (Chapter 4).
- Has the patient been tested before? Germline genotype does not change, so an earlier result may be enough, if it covered the gene and variants you need (Chapter 3).
- Who pays, and what will it cost the patient? Coverage varies by payer, contractor and indication.
Check drug-specific genetic safety requirements before prescribing
- Carbamazepine in a patient with ancestry in a population at genetic risk: the label's boxed warning says to screen for HLA-B*15:02 before starting [22]. See Chapter 2
- Carbamazepine in a patient known to carry HLA-A*31:01: the US label warns of hypersensitivity reactions with this allele, without the same ancestry-based screening instruction; weigh the risks and benefits before prescribing [22]
- Oxcarbazepine in a patient with ancestry in a genetically at-risk population: the label says HLA-B*15:02 testing should be considered before starting [23]. See Chapter 2
- Valproate in a patient with a family history or symptoms suggesting a POLG-related mitochondrial disorder: the label calls for POLG testing in line with current diagnostic practice, and valproate is contraindicated in known POLG disorders and in children under 2 suspected of having a POLG-related disorder. In patients over 2 with suspected hereditary mitochondrial disease, the label limits use to cases where other anticonvulsants have failed, with regular clinical assessments and serum liver tests [24]. See Chapter 2
Consider a test, after weighing existing results, alternatives and cost, if
- You are dosing a current drug, or choosing a new one, with a CPIC, DPWG or FDA label recommendation for a gene on the test
- Side effects at low doses, or no response at adequate dose and duration, on drugs with such recommendations
- The patient is likely to need several drugs with such recommendations over time (pre-emptive testing)
Don't order: counsel and record
The patient is doing well on a tolerated drug, the question is which drug will work, or an adequate result already exists. Explain what a result can and cannot do, and record the discussion.
Prescriber-side triage for pharmacogenomic testing. Our synthesis of the trials, guidelines, labels, coverage policies and policy statements cited in this chapter, not a validated instrument. The US labels write the allele as HLA-B*1502; this chapter uses the current notation, HLA-B*15:02.
Factors that can encourage low-value testing
- Tests are marketed directly to patients and clinicians
- Reports sort drugs into categories that look like recommendations
- Trial results are often quoted without the primary outcome
Record in the chart
- The decision the result would change
- Drug being used or considered, and its gene–drug guidance
- Prior trials, doses and outcomes
- Prior genetic test: lab, date and genes covered
- What was discussed, including cost
Look up the situation
Tap the situation for what it means and what to do.
Safety check first
Consider
Counsel
Starting carbamazepine or oxcarbazepine
Safety check firstThis is a targeted genetic safety question, distinct from broad antidepressant-selection testing. Confirm that the assay reliably covers the required allele. Chapter 2 covers the decision.
What to do
- Carbamazepine: screen for HLA-B*15:02 before the first dose in patients with ancestry in at-risk populations; don't use it in a positive patient unless the benefit clearly outweighs the risk. The US label also warns of hypersensitivity reactions, including SJS/TEN, maculopapular eruptions and DRESS, with HLA-A*31:01; in a patient known to carry it, weigh the risks and benefits before prescribing [22]
- Oxcarbazepine: consider HLA-B*15:02 testing in the same patients; avoid it in a positive patient unless the benefits clearly outweigh the risks [23]
Keep in mind
- The oxcarbazepine label does not generally recommend screening current users, because the risk is largely confined to the first few months of therapy [23]
- Both labels state that genotyping never replaces clinical vigilance; reactions can occur in HLA-negative patients [22][23]
- The 2026 UK CERSI-PGx guideline goes further than US labeling: testing regardless of ancestry for patients starting carbamazepine, oxcarbazepine or eslicarbazepine, or taking them for less than 3 months, covering HLA-B*15:02, HLA-B*15:11 and HLA-A*31:01 [21]
Side effects or no response on earlier trials
ConsiderA result can help if the current drug, or one you are considering, has a gene–drug recommendation.
What to do
- Confirm adherence, dose, duration and interacting drugs first (Chapter 4)
- Check the current or contemplated drug against CPIC and DPWG guidance [13][14][15][16]
- If it has a recommendation, a result may change its dose or choice
Keep in mind
- No trial set a number of failed trials that justifies testing. GUIDED required at least one, and its patients averaged 3.5 [3]
- Trial benefits were small and not consistent across time points [1][2]
Patient asks for “the gene test”
CounselExplain what a result can and cannot do, then decide together whether it would change the plan.
What to do
- Ask what the patient hopes the test will show
- Explain that it informs dose and drug choice for some drugs, not reliably which drug will work
- Check coverage and cost before ordering
Keep in mind
- In October 2018 the FDA warned against genetic tests with unapproved claims to predict patient response to specific medications [10]
- A result the patient already has, including a direct-to-consumer one, is covered in Chapter 3
All three situations on one page (figure)
Figure 1. Pharmacogenomic panels: what a result can and can't reliably do. Our summary of the trials and guidelines cited in this chapter [1][2][3][8][12].
What the trials actually showed
In depression trials, test-guided care changed which drugs were prescribed more reliably than it changed outcomes. GUIDED missed its primary endpoint. PRIME Care found a small improvement in remission across 24 weeks, a co-primary outcome, but no significant difference at week 24. ADOPT PGx did not meet its primary endpoint at 3 months and found higher remission at 6 months as a secondary outcome. PREPARE tested adverse drug reactions across medicine, not response to antidepressants.
| Trial | Design and funding | What it found | Source type |
|---|---|---|---|
| GUIDED (2019) | Randomized trial of a combinatorial commercial panel (GeneSight) versus treatment as usual in outpatients with major depression and an inadequate response to at least one antidepressant (mean 3.5 failed medications). Of 1,799 randomized patients, 258 were excluded after randomization, leaving 1,541 in the intention-to-treat cohort; the primary outcome was analyzed in the per-protocol cohort of 1,398, of whom 1,167 completed the blinded week-8 visit. Patients and raters blinded to week 8; treating clinicians not blinded. The paper has no funding statement; the trial registry lists Assurex Health, the test's manufacturer, as sponsor [32], and five authors were Assurex Health employees | Primary outcome, symptom improvement at week 8: 27.2% vs 24.4%, P=.107, not significant. Secondary outcomes at week 8: response 26.0% vs 19.9% (P=.013) and remission 15.3% vs 10.1% (P=.007) | Peer-reviewed randomized trial; industry-sponsored per registry [3][32] |
| PRIME Care (2022) | Pragmatic, single-blind randomized trial at 22 Veterans Affairs medical centers; 1,944 veterans with major depression starting or switching a single antidepressant; mean age 48; 25% women. Test: GeneSight, run by Myriad Genetics. Outcome raters were blinded; clinicians and patients were not. Funded by a VA Health Services Research and Development Merit Review Award; Myriad supplied genotyping at no cost; VA staff took part in the design, conduct and interpretation | Co-primary outcome 1: more prescriptions with no predicted drug–gene interaction in the first 30 days (59.3% vs 25.7%; OR 4.32, 95% CI 3.47 to 5.39). Co-primary outcome 2, remission analyzed across repeated assessments over 24 weeks: OR 1.28 (95% CI 1.05 to 1.57), risk difference 2.8 percentage points (0.6 to 5.1). Significant at weeks 8 and 12 only; at week 24, risk difference 1.5 percentage points (95% CI −2.4 to 5.3), P=.45. The authors described the effects on remission as small and nonpersistent | Peer-reviewed randomized trial [1] |
| ADOPT PGx Depression (2026) | Pragmatic randomized trial in primary care, psychiatry and family medicine clinics at 9 US health systems, August 2021 to April 2024; 1,460 patients aged 8 or older (1,239 adults, 221 children). CYP2D6- and CYP2C19-guided SSRI prescribing using nonproprietary recommendations based on CPIC guidelines, FDA labeling or both, accounting for CYP2D6-inhibiting drugs. Outcomes analyzed among the 692 patients (47.4%) with an actionable phenotype. Outcome assessors masked; clinicians and patients not masked. Funded by the NIH (IGNITE network); NIH program officers contributed to the study design [2][33] | Primary outcome, change in depression score at 3 months, not met (−4.3 vs −4.0, P=.68; covariate-adjusted P=.67). Secondary outcomes: no difference in adverse-effect burden at 3 months (P=.37); remission at 6 months 48.3% vs 39.4% (P=.02) on PROMIS and 29.9% vs 21.0% (P=.01) on PHQ-8. All results are among patients with an actionable result. The analysis plan reported secondary outcomes with unadjusted P values rather than correcting for multiple comparisons | Peer-reviewed randomized trial [2] |
| PREPARE (2023) | Open-label, cluster-randomized crossover implementation study in 7 European countries; 6,944 adults starting a drug covered by DPWG guidance, which triggered testing; 12-gene panel, with treatment guided by DPWG recommendations. Funded by the European Union's Horizon 2020 program | Clinically relevant adverse drug reactions within 12 weeks: 21.0% vs 27.7% in patients with an actionable result (OR 0.70, 95% CI 0.54 to 0.91). It did not test response to antidepressants | Peer-reviewed implementation trial [4] |
| PRIME Care reanalysis (2025) | Post hoc analysis of PRIME Care, run on an analysis plan written beforehand; 1,764 of 1,944 patients. Eight of the ten authors were Myriad Genetics employees; the trial's lead investigator reported nonfinancial support from Myriad for the conduct of the trial. The authors declared no financial support for the analysis | A different outcome from the original paper: time to first remission, not remission status at week 24. Reported earlier first remission (HR 1.27, 95% CI 1.05 to 1.53) and first response (HR 1.21) with test-guided care, with no evidence that the effect changed over 24 weeks. Earlier first remission does not establish sustained remission, and a nonsignificant test for change in the effect over time is not proof of durable remission. Myriad issued a press release and said it plans to submit the data to payers | Peer-reviewed post hoc analysis [5]; company press release, not peer-reviewed [6] |
The detail: how to read these trials
Each depression trial tested a whole package: a test, a report and advice to the clinician. A positive result for the package cannot show that every gene on the panel contributes. GUIDED and PRIME Care used a proprietary combinatorial test; in PRIME Care it combined 4 pharmacodynamic and 8 pharmacokinetic gene variants, and the PRIME Care authors noted its algorithm may not align with CPIC recommendations [1]. ADOPT PGx used two genes and publicly available CPIC and FDA prescribing guidance [2], so it is a cleaner test of acting on established drug–gene recommendations.
Meta-analyses that pool these trials have favored testing. The MolDX Medicare policy’s evidence review describes two (Rosenblat et al. 2018; Bousman et al. 2019) that found higher remission with testing, with high heterogeneity between studies, and notes that its expert advisers did not support combinatorial testing [26]. Pooled estimates depend on which trials and test platforms are included; they do not establish reliable prediction of an individual patient’s response or that one commercial panel is better than another.
None of these trials blinded the treating clinician. The PRIME Care authors noted that the modest effect could be a placebo-type effect for that reason [1]. The GUIDED paper states that treating clinicians were not blinded [3], although its registry record lists care providers as masked [32]. GUIDED analyzed its primary outcome in a per-protocol cohort [3]. In ADOPT PGx, the depression trial was registered on its own in July 2023, after enrollment began; the overarching platform trial was registered in June 2020, before enrollment [2][33]. Most ADOPT PGx patients were enrolled during the COVID-19 pandemic, and the trial did not formally confirm a diagnosis of major depression [2].
The 2025 reanalysis measured time to first remission, while the original paper concluded that the remission effect was small and nonpersistent [1][5]. Our view: prespecified outcomes from the original trial should lead the interpretation; post hoc analyses address additional questions and should be labeled as such, whoever writes them. Sponsorship and author relationships are disclosed separately in the table. When a trial is quoted to you, ask for the primary outcome at the primary time point, then the secondary outcomes.
What guideline bodies say about ordering
An American Psychiatric Association work group concluded in 2018, and again in 2024, that the evidence does not support using currently available combinatorial tools to select depression treatment. The 2022 VA/DoD depression guideline found insufficient evidence to recommend for or against testing. AACAP advises against using testing to select psychotropics in children and adolescents. All three positions predate ADOPT PGx. CPIC tells you how to use a result already in hand, not whether to order the test; DPWG also rates whether testing is potentially beneficial for some pairs. Their recommendations, and FDA labeling, can differ in which phenotypes need action and by how much, so cite the specific source.
| Body | Position | Source |
|---|---|---|
| APA Council of Research Workgroup on Biomarkers and Novel Treatments | 2018: insufficient data to support widespread use of combinatorial pharmacogenetic testing in clinical practice, although it may be informative in some situations, particularly in predicting side effects. 2024 update: of 11 newer trials, 3 met their primary outcome and all had significant limitations; the evidence does not support currently available combinatorial tools for treatment selection in major depression. The 2024 review covered trials published from 2017 to 2022, so it predates ADOPT PGx | Zeier et al. 2018 [7]; Baum et al. 2024 [8] |
| VA/DoD Management of Major Depressive Disorder guideline (February 2022) | Insufficient evidence to recommend for or against pharmacogenetic testing to help guide the selection of antidepressants (Recommendation 13, neither for nor against). The work group rated the studies it reviewed as very low quality and noted concern about bias because all were commercially funded. The guideline predates PRIME Care and ADOPT PGx | VA/DoD 2022 [35] |
| American Academy of Child and Adolescent Psychiatry | Clinicians should avoid using pharmacogenetic testing to select psychotropic medications in children and adolescents. The 2020 statement predates ADOPT PGx, which enrolled children | AACAP policy statement, 2020 [9] |
| US Food and Drug Administration | Warned in October 2018 against genetic tests with unapproved claims to predict patient response to specific medications, including some used for depression. Its Table of Pharmacogenetic Associations states that listing a gene–drug pair does not necessarily mean the FDA advocates testing before prescribing | FDA safety communication, 2018 [10]; FDA table [11] |
| CPIC | Its guidelines assume the test results are in hand and do not discuss the merits of doing the test or its cost-effectiveness. It gives recommendations for using a result, for example for SSRIs, SNRIs and vortioxetine. Its CYP2D6–antipsychotic guideline is still in progress | CPIC standard operating procedure, 2025 [12]; Bousman et al. 2023 [13]; CPIC website [19] |
| DPWG | Rates genotyping for some pairs, such as CYP2D6 with venlafaxine, as potentially beneficial, to be considered for individual patients rather than all patients; the same applies to its antipsychotic pairs. Its venlafaxine advice (avoid in poor and intermediate metabolizers) is stronger than CPIC’s (an optional alternative in poor metabolizers only) | DPWG 2024 guidelines [14][15]; Bousman et al. 2023 [13] |
Who a result may help
Testing has its clearest value when a plausible result would change a specific prescribing decision. A result may support either standard prescribing or a drug-specific adjustment; neither establishes which antidepressant will produce remission. It does not help you choose between drugs that have no gene–drug recommendation.
| Situation | Why a result may help, or not | Chapter |
|---|---|---|
| Starting, continuing or dosing an antidepressant with a CPIC or DPWG recommendation | CPIC 2023 covers citalopram, escitalopram, fluvoxamine, paroxetine, sertraline (CYP2C19 and, in CPIC only, CYP2B6), venlafaxine and vortioxetine [13]; TCAs have CPIC and 2026 DPWG guidance [16][17]. For the phenotypes the guideline names, a result can change the starting dose, titration or drug choice | 5 and 6 |
| Starting or dosing an antipsychotic with a DPWG recommendation | DPWG recommends adjustments for specified CYP2D6 phenotypes with aripiprazole, brexpiprazole, haloperidol, pimozide, risperidone and zuclopenthixol, and for CYP3A4 poor metabolizers with quetiapine [14] | 5 and 7 |
| Starting atomoxetine | CPIC has a CYP2D6 guideline for atomoxetine [18] | 5 |
| Side effects at low doses, or no response at adequate dose and duration | May reflect altered metabolism; can also reflect adherence, trial length or an interacting drug. Helps if the current or a contemplated drug has a recommendation | 4 |
| Likely to need several drugs with recommendations over time | Pre-emptive testing means the result is ready when needed. PREPARE supports panel-guided prescribing across medical specialties when an eligible drug is being started; it did not establish the benefit of population-wide psychiatric screening or reliable prediction of antidepressant response [4] | 3 |
| Choosing among drugs without a recommendation | A result adds little. DPWG found no need for adjustment for CYP2D6 with duloxetine and CYP2C19 with mirtazapine [15], or for CYP1A2 with clozapine and olanzapine [14] | 7 |
The detail: reactive versus pre-emptive testing
Reactive testing is ordered for a decision in front of you, such as which dose of a drug to start. Pre-emptive testing is ordered so that a result is on file before it is needed; CPIC notes that its recommendations apply to results generated either way [12]. Our view: in psychiatry, reactive testing tied to a named drug is easier to justify, because the result has a specific use. Pre-emptive testing makes most sense where the result will be stored where future prescribers will see it, and where the patient is likely to need drugs with recommendations.
The detail: a targeted test or a panel?
Start with the gene–drug question, then choose an assay that covers the variants that question needs. A panel adds information for possible future prescriptions, but adding genes does not by itself add clinical utility, and results from one company’s reporting system cannot be assumed for another’s. The MolDX Medicare policy reflects this: it covers a multi-gene panel when more than one gene on it is needed for the drug or drugs being considered, but not when only one gene is, and it does not cover the proprietary algorithms in combinatorial tests [26].
Does this apply to my patient?
The trials above mostly enrolled adults, and CPIC notes that its antidepressant recommendations come largely from studies of people of European or East Asian ancestry. These groups raise different questions: whether the assay covers the variants common in the patient’s ancestry, whether the prescribing evidence applies to the patient, and whether nongenetic factors such as pregnancy or interacting drugs change actual drug exposure. A normal result from a limited assay does not exclude an untested functional variant.
| Patient | What changes | Source |
|---|---|---|
| Children and adolescents | AACAP advises against using testing to select psychotropics [9]. This is a different question from dosing a drug already chosen. CPIC states that citalopram, escitalopram and sertraline have the most pharmacogenetic data in children, so its recommendations for these drugs are relevant to pediatric patients; because CYP2D6, CYP2C19 and CYP2B6 activity reaches adult levels by early childhood, extending its other antidepressant recommendations to adolescents, or possibly younger children, may be appropriate with close monitoring [13]. A published response, whose authors include CPIC guideline authors, argues that AACAP's all-or-nothing position misses such dosing uses [20]. ADOPT PGx enrolled 221 children aged 8 to 17, but its main paper reports children and adults together [2] | AACAP 2020 [9]; CPIC 2023 [13]; Ramsey et al. [20]; ADOPT PGx [2] |
| Ancestry | In PREPARE, 97.7% of patients self-reported European, Mediterranean or Middle Eastern ethnicity [4]; in ADOPT PGx, 65.6% were White or European American [2]. CPIC notes there is no reason to expect the CYP effects to differ by ancestry, but its evidence comes mainly from European and East Asian populations [13]. Panels differ in which variants they test, and limited coverage can miss reduced- or no-function alleles and produce an incorrect normal-function call. A study of Australian panels gave CYP2D6*29 as an example: about 0.1% in European ancestry but 8.7% to 10.8% in African ancestry [28]. One Medicare contractor requires reports to say when a normal (reference) allele was assigned because no other allele was found [26] | PREPARE [4]; ADOPT PGx [2]; CPIC 2023 [13]; Forbes et al. 2023 [28]; MolDX LCD [26]; Chapter 3 |
| Pregnancy | Genotype does not change in pregnancy, but enzyme activity does: CYP2D6 and CYP3A4 activity increases and CYP1A2 and CYP2C19 activity decreases [29]. Plasma SSRI concentrations vary considerably during pregnancy, partly for this reason [30]. Clinical response and, where available, drug levels can matter more than the genotype prediction | Pariente et al. 2016 [29]; Poweleit et al. 2022 [30]; Chapter 4; Perinatal Psychopharmacology course |
| Older adults on several drugs | A systematic review found that CYP-inhibiting drugs, increasing age, cancer and inflammation can make patients metabolize drugs more slowly than their genotype predicts; its clinical impact remains unclear [31]. Review the full medication list before acting on a result | Klomp et al. 2020 [31]; Chapter 4 |
Cost, coverage and retesting
Check coverage before ordering. Medicare coverage is set by contractor-specific local policies, and the ones reviewed here cover testing only when a named drug with an actionable gene–drug interaction is being considered or already used. Germline genotype does not change, so an adequate earlier result usually needs reinterpretation, not repetition.
| Question | Answer | Source |
|---|---|---|
| When does Medicare cover it? | Medicare's national coverage determination on pharmacogenomic testing (NCD 90.1) addresses warfarin response only [25]; other testing is governed by local coverage determinations, which differ by contractor. The MolDX policy (L38335, version effective 5 March 2026) covers testing once the clinician has narrowed treatment to specific drugs, or is already using one, that have a gene–drug interaction made actionable by the FDA label or CPIC level A or B. It does not cover testing on the basis of a diagnosis alone, genes without an actionable use, or the proprietary algorithms in combinatorial tests [26]. Another contractor's policy (L39063) uses similar criteria [34]. Check which policy applies to your patient's contractor | NCD 90.1 [25]; L38335 [26]; L39063 [34] |
| Does commercial insurance cover it? | Coverage varies by insurer and indication. Check before ordering, and tell the patient the likely out-of-pocket cost | Our advice |
| Should I repeat a test? | Usually not to recheck the same genes. L39063 covers germline testing once in a lifetime and treats repeat testing of the same content as duplicative [34]. But panels differ in which variants they test; in the Australian study, no two commercial panels tested the same set [28]. Expanded or confirmatory testing can be justified when an earlier report did not cover the gene or variant you need, or its result is not reliable | L39063 [34]; Forbes et al. 2023 [28]; Chapter 3 |
| Can I trust the report's categories? | Not without checking. In a retrospective review of 100 patients, 50 tested by each of two companies, 32 of 250 genotype-to-phenotype comparisons (12.8%) differed from CPIC, all from one company. Using the investigators' CPIC-based binning method, 114 of 266 medication recommendations (42.9%) differed from CPIC-derived categories [27]. The study compared how the companies interpreted genotypes and turned them into recommendations; it did not retest the DNA | Nguyen et al. 2024 [27]; Chapter 3 |
| The patient has a direct-to-consumer result | Read it before ordering anything new; Chapter 3 covers when it can be used | Chapter 3 |
Clinical pearls
- Exposure and safety, not reliable individual remission prediction.CYP results inform drug exposure and selected variants inform safety; neither reliably identifies the antidepressant that will produce remission.
- Name the decision first.If the result won't change it, don't order.
- Ask for the primary outcome.GUIDED missed it. PRIME Care found a small improvement in remission across repeated assessments over 24 weeks, but no significant difference at week 24.
- Check who ran the analysis.Industry sponsorship or authorship doesn't make a result wrong, but it should be stated.
- Genotype is for life.Find the old report and check what it covered before ordering a new one.
- Safety checks are a separate question.HLA before carbamazepine is Chapter 2. Confirm the assay covers the allele.
- Children: AACAP advises against using tests to choose drugs.
Patient script & EMR note
Counseling script: when a patient asks for the gene test
"This test looks at genes that affect how your body breaks down some medicines. For certain medicines, it can help me choose the dose, or steer away from one your body may handle poorly.
What it can't reliably do is tell us which medicine will work for you. In the large studies, the test changed which medicines people were given more than it changed how well they did.
If the medicine we're using or considering is one where the result would change what I do, it may be worth doing. Before we order it, let's check what your insurance covers, because the cost varies. Your genes don't change, so if you've had a test like this before, bring the report and we'll check whether it covers what we need."
EMR note: copy, paste and edit
PHARMACOGENOMIC TESTING – DECISION
Decision the result would change: [ ]
Drug(s) used or being considered: [ ]
CPIC/DPWG/FDA label recommendation
for drug: [Y/N] Gene(s): [ ]
Prior trials (drug, dose, duration,
outcome): [ ]
Low-dose intolerance / nonresponse: [ ]
Interacting drugs: [ ]
Prior genetic test: [Y/N] Lab/date: [ ]
Covers needed genes/variants: [Y/N]
Applies: [child/adolescent, pregnancy,
older adult, ancestry]
Coverage checked: [Y/N]
Discussed:
[ ] May inform dose/drug choice; does not
reliably predict individual response
[ ] SLC6A4/HTR2A/ANK3/MTHFR not used to
select or dose
[ ] Cost
Plan:
[ ] Targeted test: [ ]
[ ] Panel: [ ]
[ ] Not ordered. Reason: [ ]
[ ] Review on: 2026
Clinician: [ ] Date/time: [ ]
Case: Thursday, 2:00 pm
A 29-year-old woman with major depression stopped escitalopram after two weeks because of nausea and jitteriness at the starting dose. She has now taken sertraline for eight weeks with only a partial response. She takes no other medicines and is not pregnant. She has read about “the gene test” online and hopes it will show which antidepressant will work for her.
By the end of this chapter you should be able to say whether to order a test and what to tell her it can and cannot show.
Case resolution
Her current drug, sertraline, has CPIC recommendations based on CYP2C19 and CYP2B6, and escitalopram has CYP2C19 recommendations [13], so a result could identify an exposure-related factor relevant to dosing. That makes testing an option for her, not a requirement. If the next choice is a drug without a recommendation, such as mirtazapine for CYP2C19 [15], the result would add little. Her clinician first checks adherence and whether the sertraline dose has been optimized, then explains that the trials showed small and inconsistent benefits [1][2][3] and that no test reliably predicts which drug will work. They discuss the options that do not depend on a test: optimizing the sertraline dose if it is not yet adequate, switching treatment, or adding evidence-based psychotherapy. Testing remains an option if a result would change the dose or drug choice, after checking her coverage. Our view: treatment should not wait for an optional panel, and the discussion and the decision belong in the chart.
Why it matters
Pharmacogenomic panels are marketed widely, and patients increasingly ask for them. A result can change the dose or choice of some commonly prescribed psychiatric drugs, and in those cases it is useful. But trial results are often quoted without their primary outcomes, reports can look more definitive than the evidence behind them, and the FDA has warned about tests that claim to predict response [10]. Knowing what a result can and cannot do lets you order the test when it will change a decision and decline it when it will not.
References (35)
- Oslin DW, Lynch KG, Shih MC, 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. doi:10.1001/jama.2022.9805. Source type: peer-reviewed randomized trial.
- Blake KV, Hines LJ, Liu M, et al. Genotype-guided antidepressant prescribing for patients with depression: a randomized clinical trial. JAMA Netw Open. 2026;9(5):e2610609. doi:10.1001/jamanetworkopen.2026.10609. Including Supplement 1 (trial protocol and statistical analysis plan). Source type: peer-reviewed randomized trial.
- Greden JF, Parikh SV, Rothschild AJ, et al. Impact of pharmacogenomics on clinical outcomes in major depressive disorder in the GUIDED trial: a large, patient- and rater-blinded, randomized, controlled study. J Psychiatr Res. 2019;111:59–67. doi:10.1016/j.jpsychires.2019.01.003. Source type: peer-reviewed randomized trial; registry sponsor Assurex Health [32]; five authors employed by Assurex Health.
- Swen JJ, van der Wouden CH, Manson LEN, et al. A 12-gene pharmacogenetic panel to prevent adverse drug reactions: an open-label, multicentre, controlled, cluster-randomised crossover implementation study. Lancet. 2023;401(10374):347–356. doi:10.1016/S0140-6736(22)01841-4. Source type: peer-reviewed implementation trial.
- Hain D, Del Tredici AL, Griggs RB, et al. Persistent benefit of pharmacogenomic testing on initial remission and response rates in patients with major depressive disorder. Front Pharmacol. 2025;16:1658616. doi:10.3389/fphar.2025.1658616. Source type: peer-reviewed post hoc analysis; eight of ten authors employed by Myriad Genetics.
- Myriad Genetics. New post-hoc analysis shows patients whose clinicians had access to GeneSight results for depression treatment are more likely to feel better sooner. Press release, GlobeNewswire, 31 October 2025. Source type: company press release, not peer-reviewed.
- Zeier Z, Carpenter LL, Kalin NH, et al. Clinical implementation of pharmacogenetic decision support tools for antidepressant drug prescribing. Am J Psychiatry. 2018;175(9):873–886. doi:10.1176/appi.ajp.2018.17111282. Source type: professional society work group review.
- Baum ML, Widge AS, Carpenter LL, et al.; APA Workgroup on Biomarkers and Novel Treatments. Pharmacogenomic clinical support tools for the treatment of depression. Am J Psychiatry. 2024;181(7):591–607. doi:10.1176/appi.ajp.20230657. Erratum: Am J Psychiatry. 2024;181(12):1134. Abstract read. Source type: professional society work group review.
- American Academy of Child and Adolescent Psychiatry. Clinical use of pharmacogenetic tests in prescribing psychotropic medications for children and adolescents. Policy statement, approved by Council March 2020. Accessed October 2026. Source type: professional society policy statement.
- US Food and Drug Administration. The FDA warns against the use of many genetic tests with unapproved claims to predict patient response to specific medications: FDA safety communication. Issued 31 October 2018; updated 4 April 2019. Read via an archived copy; the original page is no longer online. Source type: regulatory communication.
- US Food and Drug Administration. Table of Pharmacogenetic Associations. Accessed 25 October 2026. Source type: regulatory reference table.
- Clinical Pharmacogenetics Implementation Consortium. Clinical practice guideline standard operating procedure. Revised 9 May 2025. Source type: guideline methods document.
- Bousman CA, Stevenson JM, Ramsey LB, 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. doi:10.1002/cpt.2903. Source type: peer-reviewed guideline.
- Beunk L, Nijenhuis M, Soree B, et al. Dutch Pharmacogenetics Working Group (DPWG) guideline for the gene-drug interaction between CYP2D6, CYP3A4 and CYP1A2 and antipsychotics. Eur J Hum Genet. 2024. doi:10.1038/s41431-023-01347-3. Source type: peer-reviewed guideline.
- Beunk L, Nijenhuis M, Soree B, et al. Dutch Pharmacogenetics Working Group (DPWG) guideline for the gene-drug interaction between CYP2D6, CYP2C19 and non-SSRI/non-TCA antidepressants. Eur J Hum Genet. 2024;32:1371–1377. Source type: peer-reviewed guideline.
- Kleine Schaars K, Nijenhuis M, Soree B, et al. Dutch Pharmacogenetics Working Group (DPWG) guideline for the gene-drug interaction between CYP2D6 and CYP2C19 and tricyclic antidepressants. Eur J Hum Genet. 2026;34(3):379–386. doi:10.1038/s41431-025-02008-3. Source type: peer-reviewed guideline.
- Hicks JK, Sangkuhl K, Swen JJ, et al. Clinical Pharmacogenetics Implementation Consortium guideline (CPIC) for CYP2D6 and CYP2C19 genotypes and dosing of tricyclic antidepressants: 2016 update. Clin Pharmacol Ther. 2017;102(1):37–44. doi:10.1002/cpt.597. Source type: peer-reviewed guideline.
- Brown JT, Bishop JR, Sangkuhl K, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2D6 genotype and atomoxetine therapy. Clin Pharmacol Ther. 2019;106(1):94–102. doi:10.1002/cpt.1409. Source type: peer-reviewed guideline.
- Clinical Pharmacogenetics Implementation Consortium. Guidelines published, in progress or planned (CYP2D6 and antipsychotics: in progress). cpicpgx.org. Accessed October 2026. Source type: guideline consortium website.
- Ramsey LB, Namerow LB, Bishop JR, et al. Thoughtful clinical use of pharmacogenetics in child and adolescent psychopharmacology. J Am Acad Child Adolesc Psychiatry. 2021;60(6):660–664. doi:10.1016/j.jaac.2020.08.006. Source type: peer-reviewed letter.
- Galloway L, Dello Russo C, Bass N, et al. HLA genotype testing for carbamazepine, oxcarbazepine and eslicarbazepine: a guideline developed by the UK Centre of Excellence in Regulatory Science and Innovation in Pharmacogenomics (CERSI-PGx). Br J Clin Pharmacol. 2026;92(7):1957–1976. doi:10.1002/bcp.70559. Full text read. Source type: peer-reviewed guideline (UK).
- Tegretol (carbamazepine) [US prescribing information]. Boxed warning; Warnings. Label text revised July 2026. Accessed October 2026 via DailyMed. Source type: US product label.
- Trileptal (oxcarbazepine) [US prescribing information]. Warnings and precautions. Revised September 2025. Accessed October 2026 via DailyMed. Source type: US product label.
- Depakote (divalproex sodium) [US prescribing information]. Contraindications; warnings and precautions. Revised March 2026. Accessed October 2026 via DailyMed. Source type: US product label.
- Centers for Medicare & Medicaid Services. National coverage determination 90.1: Pharmacogenomic testing for warfarin response. Accessed October 2026. Source type: US payer coverage policy.
- Centers for Medicare & Medicaid Services. Local coverage determination L38335: MolDX: Pharmacogenomics Testing. Version 16; revision effective 5 March 2026. Accessed October 2026. Source type: US payer coverage policy.
- Nguyen TT, Leary EJW, Lee JT, Shukla SK, Griesbach SA. Comparing commercial pharmacogenetic testing results and recommendations for antidepressants with established CPIC guidelines. Front Pharmacol. 2024;15:1500235. doi:10.3389/fphar.2024.1500235. Source type: peer-reviewed retrospective chart review.
- Forbes M, Hopwood M, Bousman CA. CYP2D6 and CYP2C19 variant coverage of commercial antidepressant pharmacogenomic testing panels available in Victoria, Australia. Genes (Basel). 2023;14(10):1945. doi:10.3390/genes14101945. Source type: peer-reviewed survey of laboratory panels (Australian, not US).
- Pariente G, Leibson T, Carls A, et al. Pregnancy-associated changes in pharmacokinetics: a systematic review. PLoS Med. 2016;13(11):e1002160. doi:10.1371/journal.pmed.1002160. Source type: peer-reviewed systematic review.
- Poweleit EA, Cinibulk MA, Novotny SA, et al. Selective serotonin reuptake inhibitor pharmacokinetics during pregnancy: clinical and research implications. Front Pharmacol. 2022;13:833217. doi:10.3389/fphar.2022.833217. Source type: peer-reviewed perspective with simulation.
- Klomp SD, Manson ML, Guchelaar HJ, Swen JJ. Phenoconversion of cytochrome P450 metabolism: a systematic review. J Clin Med. 2020;9(9):2890. doi:10.3390/jcm9092890. Source type: peer-reviewed systematic review.
- ClinicalTrials.gov. Genomics Used to Improve DEpression Decisions (GUIDED). NCT02109939. Lead sponsor: Assurex Health Inc. Accessed October 2026. Source type: trial registry record.
- ClinicalTrials.gov. A Depression and Opioid Pragmatic Trial in Pharmacogenetics (Depression Trial). NCT05966155. Lead sponsor: Duke University; collaborator: National Human Genome Research Institute. First submitted 21 July 2023. Accessed October 2026. Source type: trial registry record.
- Centers for Medicare & Medicaid Services. Local coverage determination L39063: Pharmacogenomics Testing. Version 11; original effective date 12 December 2021. Accessed October 2026. Source type: US payer coverage policy.
- US Department of Veterans Affairs and Department of Defense. VA/DoD clinical practice guideline for the management of major depressive disorder. Version 4.0. February 2022. Recommendation 13, pp. 41–42. Source type: government clinical practice guideline.
Last reviewed October 2026. Part of the Psychiatry Education Forum Academy, Genetic Testing in Psychiatry · Section A · Chapter 1.
Next: Chapter 2
Targeted genetic safety assessment before selected medications: when testing is required, recommended or considered, and how the result changes prescribing.
This chapter is educational material for licensed prescribers and prescribers in training. It is not a substitute for clinical judgment, local protocols, the current prescribing information, the laboratory's report or the patient's insurance coverage policy. Psychiatry Education Forum Academy has no financial relationship with any pharmacogenomic testing company. Psychiatry Education Forum Academy is not a CME provider and this chapter does not carry CME credit.
Responses