Pharmacogenetic Decision Support Tools: A Component of Precision Medicine for Psychiatry?

If there is one overriding vision for the future of psychiatric practice, it is “precision medicine,” also referred to as “personalized medicine.” Building from successes in other medical fields, this approach aims to move from treatment algorithms based on average outcomes in clinical trials toward treatment selections tailored to each patient on the basis of measurement of specific characteristics (1). Under this model, the current trial-and-error approach to treatment would be improved by selecting treatments expected to have a greater likelihood of benefit or improved tolerability. Over the past five years, precision medicine studies in psychiatry have increased substantially, with positive results reported for neuroimaging, electroencephalography, genetics, inflammation, and clinical methods (2).

To date, none of these approaches have been replicated sufficiently to warrant incorporation into clinical practice. However, a market has developed for pharmacogenetic (PGx) decision support tools, which offer the promise of providing genetically based information to aid medication selection and dosing for individual patients across a variety of psychiatric disorders. Over 20 companies around the world currently market PGx tests (3), and some insurance companies are partially covering the tests’ costs. Patients are asking about these tests with increasing frequency, leaving prescribers with questions of whether PGx testing is justified and if so, at what point in treatment PGx should be used, and which company’s test to choose.

The PGx companies’ test reports present data in a variety of formats, but most provide both the genetic results of the specific alleles tested and the implications of these results for determining the patient’s metabolic phenotype (for pharmacokinetic genes) or likely implications for drug response (for pharmacodynamic genes). Finally, using proprietary algorithms, the reports present the clinically actionable results: lists of medications that can be used without concern (i.e., standard usage) and those that require caution or dosing outside the normal expected ranges. The reports are careful not to say “Do not use,” but the warning colors (yellows and reds) applied to the medications listed as requiring caution certainly raise a challenge to convince a patient that a medication so labelled should be tried.

The published data on these PGx tests cannot yet be considered to conclusively prove that they have clinical value (4). Many companies have no published data on the utility of their tests for psychiatric conditions. Only two small randomized trials have been published to date; both reported increased remission rates for the PGx-informed treatment arm over the treatment-as-usual arm (5, 6), with one achieving statistical significance (6). Sequential cohort studies indicating improved outcomes with PGx-guided treatment have also been published, although these designs are vulnerable to substantial confounding (3).

This should be a watershed year for determining the value of PGx testing. Two companies, Assurex Health and Neuropharmagen, have completed large, parallel group randomized trials for major depressive disorder, expected to be published this year, which should clarify and quantify the value of these companies’ PGx tools. It will be important for prescribers to assess the key clinical outcomes reported that could justify the inclusion of PGx testing into clinical practice. Specifically, was the remission rate improved in the PGx-informed arm? Was the dropout rate lower in the PGx-informed arm, and was it associated with lower rates of dropout due to side effects? Was the rate of comedication (e.g., with sedatives, anxiolytics, or stimulants) lower in the PGx-informed arm?

For these outcomes, it will be important to understand the size of the effects, particularly the number needed to test in order to achieve one improved patient outcome. Furthermore, identifying any interactions between treatment outcomes and the level of treatment resistance (number of prior treatment failures) among the study participants may inform the point in treatment that PGx testing may best fit. It should be kept in mind that drug-drug interactions can complicate interpretation of many drug-gene interaction effects and are often not included in PGx testing reports.

Even if the PGx tools demonstrate value, many important questions will remain, including which company’s test a prescriber should use. Each company’s test evaluates a different constellation of genes, although all include characterizing the P450 CYP2D6 and CYP2C19 genes responsible for metabolizing many psychiatric drugs. However, even when the tests include the same genes, they are not necessarily testing the same loci within each gene. It is highly likely that by characterizing different genetic loci, these tests will produce different genetic phenotypes (e.g., characterize the metabolic status for a certain enzyme differently in the same patient), and consequently report different medication recommendations. It is also highly likely that large head-to-head comparison trials of PGx decision-support tools will never be performed. Thus, marketing and market forces will probably determine which PGx tools will remain available for prescribers to choose from.

A final word about PGx decision support tools: the reports can be seductive, to prescribers and patients alike. In a clinic office, all a prescriber can currently use to justify treatment recommendations are the published treatment guidelines, along with an appeal to the authority of their clinical experience. Average outcomes and subjective provider beliefs can pale in significance when contrasted with the apparent objectivity and specificity of a PGx report designed precisely for that individual. Of course, the practice of psychiatry involves consideration of many factors beyond the biological. If the PGx tools prove to have clinical utility, psychiatry shall need to develop ways of discussing the results with patients so that they can usefully inform, rather than dictate, treatment recommendations (7).