Pharmacogenetics studies how inherited DNA variants can modify a drug’s pharmacokinetics or, less commonly, its pharmacodynamics. In the treatment of depression, the most established clinical application concerns certain gene-drug pairs in which genotype helps predict the activity of metabolic enzymes, particularly CYP2D6, CYP2C19 and, for some drugs, CYP2B6. The result can therefore help interpret unexpected concentrations, adverse effects, or excessively low or high exposure and can guide the choice or dosing of specific antidepressants. It is not, however, a test capable of identifying in advance, with certainty, which antidepressant will produce remission in an individual patient.
This distinction is essential because the commercial term “pharmacogenomic test” encompasses very different products. CPIC guidelines describe how to use a genetic result when it is available for specific gene-drug interactions; by themselves, they do not state that testing should be routinely prescribed to everyone with depression. In randomized trials, pharmacogenomic testing has generally promoted prescribing that is more consistent with predicted gene-drug interactions, whereas the magnitude and persistence of benefit for symptoms or remission have been more modest and variable. Pharmacogenetics should therefore be integrated with diagnosis, treatment history, comorbidities, interactions, adherence, patient preferences, and clinical monitoring.
The interpretive process does not consist of detecting a single “good gene” or “bad gene.” For each locus, allelic variants and, when necessary, rearrangements or copy-number variations are identified; the combination of alleles constitutes the diplotype. Each allele is assigned, according to standardized gene-specific systems, normal, decreased, absent, or increased function. The diplotype is then used to derive a metabolic phenotype, meaning a prediction of enzymatic activity that, depending on the gene, may be classified as poor, intermediate, normal, rapid, or ultrarapid.
For CYP2D6, genotyping is technically complex because the locus contains numerous variants, deletions, duplications, multiplications, and hybrid alleles. A panel that does not adequately detect copy-number variations or variants relevant to the population being tested may assign an incomplete phenotype. For CYP2C19, both loss-of-function alleles and the increased-function allele that contributes to rapid and ultrarapid phenotypes are clinically relevant. Clinical significance nevertheless remains drug-dependent: the same reduction in enzymatic activity may increase exposure to an active drug, have limited impact when alternative metabolic pathways exist, or have different consequences when active metabolites are present.
The 2023 CPIC recommendations for serotonin reuptake inhibitors include specific associations with CYP2D6, CYP2C19, and CYP2B6. Particularly relevant examples include citalopram and escitalopram with CYP2C19, paroxetine with CYP2D6, and sertraline with CYP2C19 and CYP2B6. For tricyclic antidepressants, CYP2D6 and CYP2C19 may complementarily influence biotransformation and drug and metabolite concentrations; in this context, pharmacogenetic guidelines can, when appropriate, be integrated with therapeutic drug monitoring of plasma concentrations.
The predicted genetic phenotype does not necessarily coincide with the metabolic phenotype observed during treatment. Potent enzyme inhibitors can cause phenoconversion, making a person genetically classified as a normal metabolizer behave as a functionally intermediate or poor metabolizer. Enzyme induction, age, hepatic and renal function, inflammation, smoking for enzymes sensitive to it, comorbidities, and polypharmacy can also modify exposure. A genetic report therefore remains stable over time, but its clinical interpretation must be updated whenever the pharmacological or biological context changes.
Some commercial panels also include variants in pharmacodynamic genes, including SLC6A4 and HTR2A, or other proposed markers of antidepressant response. The presence of statistical associations in genetic studies does not automatically mean that a marker is robust enough to guide individual prescribing. In the 2023 CPIC guideline, evidence for SLC6A4 and HTR2A is considered insufficient to support clinical prescribing recommendations. Likewise, many genes proposed in proprietary panels lack independently validated recommendations comparable with those available for certain cytochrome interactions.
Combinatorial tests integrate multiple variants through proprietary algorithms and often return summary categories, sometimes displayed with colors or labels such as “favorable use,” “use with caution,” or “significant interaction.” These categories do not constitute a uniform international standard: different panels may analyze different genes, assign different weights to the same variants, and classify the same drug differently. Clinicians should therefore distinguish the primary genetic data and interactions supported by independent guidelines from the laboratory’s proprietary interpretation.
Antidepressant response is a complex, polygenic phenotype influenced by severity and clinical subtype, psychiatric and medical comorbidities, environmental exposures, adherence, pharmacokinetics, expectations, psychosocial support, and numerous biological factors not captured by current panels. A test therefore cannot demonstrate that one drug “will work” or another “will not work.” Rather, it may change the probability that a particular drug exposure is appropriate or flag a potentially relevant interaction.
The best-known clinical trials have mainly evaluated combinatorial panels. In the GUIDED trial, the primary endpoint of mean symptom improvement was not significantly greater with guided prescribing, whereas some secondary endpoints, including response and remission, favored the guided group; subsequent analyses suggested a more evident benefit among patients who at baseline were receiving drugs with predicted gene-drug interactions, but these analyses must be interpreted in light of their post hoc nature. In the pragmatic PRIME Care trial, conducted in the Veterans Affairs system, testing reduced prescribing of drugs with predicted gene-drug interactions; the remission benefit was small over the overall period and was not persistent at the final 24-week assessment.
Meta-analyses of randomized trials generally report a modest increase in the probability of response or remission at some time points, but with heterogeneity among platforms, algorithms, populations, and study designs. Some analyses show a clearer signal at 8 to 12 weeks and not at very early or later assessments. In addition, a substantial portion of the literature was generated using specific commercial panels, and transferability of results from one algorithm to another cannot be assumed. The evidence therefore supports selective clinical utility, not replacement of guideline-based treatment with a DNA-determined drug choice.
A further methodological issue is that trials may simultaneously demonstrate two different phenomena: greater adherence to pharmacogenetic recommendations and improvement in clinical outcomes. The first is directly linked to genetic information; the second also depends on prescribing quality, follow-up intensity, treatment changes, patient expectations, and other aspects of care. Correct interpretation therefore requires avoiding the transformation of an average group association into a deterministic prediction for an individual.
Pharmacogenetics is most informative when there is a defined clinical question and the drug under consideration has a validated gene-drug recommendation. It may be particularly useful if a patient has already experienced adverse effects disproportionate to usual doses, unexpected plasma concentrations, apparent lack of exposure not explained by poor adherence, or when choosing among alternatives for which the genetic result has different implications. A result already available can also be reconsidered whenever a relevant drug is prescribed.
There is no rationale for indiscriminately requiring an extensive panel for every patient as a prerequisite to starting antidepressant therapy. The absence of relevant variants does not eliminate the risk of adverse effects, and the presence of a particular metabolic phenotype does not by itself imply that the drug is contraindicated: in many situations the recommendation is to modify the dose, choose an alternative, or intensify monitoring, depending on the drug and strength of evidence.
Before using a result, it is necessary to know which alleles and copy-number variants are actually detected, the analytical method, the laboratory’s ability to distinguish structural variants, and the population in which the panel was validated. Allele frequencies vary among ancestral populations; a panel designed with limited coverage may label an untested rare allele as “wild type.” Analytical quality is therefore distinct from clinical validity: even a perfectly measured genotype may have little utility when no clinically actionable gene-drug association exists.
The discussion with the patient should clarify that a genetic result is not a psychiatric diagnosis and does not measure depression severity. Issues of informed consent, storage of genetic data, the possibility of incidental findings if the panel extends beyond pharmacogenetics, access to the data, and reuse should also be addressed. Because the germline genotype does not change, a well-documented report can avoid unnecessary repeat testing; what changes is its interpretation in light of new drugs, new guidelines, and new knowledge.
In summary, pharmacogenetics has a real but circumscribed role in the treatment of depression: it can improve prescribing precision for certain interactions, particularly metabolic ones, but it is not a “genetic map” of antidepressant response. Its greatest value is achieved when the result is used as one component of a structured and verifiable clinical decision.
Informational notice: the information contained on this page is provided solely for informational and educational purposes and does not replace the advice, diagnosis or treatment provided by a physician. If needed, always consult a qualified healthcare professional.
Artificial intelligence transparency: this page was created with the support of artificial intelligence tools, used to assist in the production and processing of its content.