Antidepressant prescribing in the precision medicine era: a prescriber’s primer on pharmacogenetic tools

Pharmacogenetic decision support tools have evolved rapidly over the past decade. In 2004, the first-generation of clinical pharmacogenetic tools was born when Roche’s CYP2D6 and CYP2C19 Amplichip (Basel, Switzerland) was made available. First-generation tools test individual genes/variants and provide genotype and accompanying phenotype (e.g. metaboliser status) information. However, they do not account for potential synergies between genetic variants and may not offer clinical interpretation/recommendations or flag drug-drug interactions. Although individual gene tests remain available for clinical use, second-generation tools now utilise combinatorial or polygene testing. The combinatorial/polygene approach is based on evidence that most antidepressants and other psychiatric medications interact with multiple pharmacodynamic and pharmacokinetic pathways [16]. Thus, unlike their first-generation predecessors, second-generation tools [see recent reviews: [10, 17] account for synergies between genes included in their testing panels and often provide drug-drug interaction information to aide in drug selection and/or dosing decisions.

Ideally, pharmacogenetic-based decision support tools would include information from a wide variety of genomic, personal, and environmental factors implicated in drug response and toxicity variability, yet current tools typically only include genetic and sometimes drug-drug interaction information (Fig. 1). This is an issue since many of the most robust predictors of response are clinical and psychosocial in nature [18]. The genetic content varies considerably from tool to tool, although what is consistent across all antidepressant pharmacogenetic tools is a focus on pharmacokinetic genes, specifically CYP2D6 and CYP2C19 [10]. The focus on CYP2D6 and CYP2C19 is primarily a result of expert groups such as the Clinical Pharmacogenetics Implementation Consortium [19] that have developed dosing guidelines for serotonin selective reuptake inhibitors and tricyclic antidepressants based exclusively on CYP2D6 and CYP2C19 genetic variation [20‐22]. Despite these guidelines, a majority of tools also include other pharmacokinetic and/or pharmacodynamics genes in their testing panels with varying degrees of evidence [10]. These gene panels are then subjected to the tool’s decision algorithm, which in turn produces an interpretative report. Interpretative reports vary in the depth of content but at minimum include a snapshot of the patient’s pharmacogenetic status along with recommendations and/or considerations aimed at optimising efficacy and/or reducing adverse events associated with antidepressant therapy.

The menu of tools available to clinicians depends on geography (Fig. 2). Clinicians in the United States have the largest selection of tools to choose from, although the availability of tools in other countries and regions of the world is growing. As a result, pharmacogenetic testing for psychotropic medication use is increasing among physicians in the United States [23] and Canada [24]. However, there is less support for direct-to-consumer genetic testing and three-quarters of US psychiatrists believed genetic counselling would be needed for patients having testing [23]. A recent study reported that 6% of psychiatrists in the United States ordered a pharmacogenetic test in the past six months, representing 47% of all genetic tests ordered by psychiatrists [25]. In other parts of the developed world, similar studies have not yet been conducted but there is no reason to think the opinions and rates of genetic testing are likely to differ. However, it remains unclear at what stage of treatment these tests are ordered or what underlying clinical circumstances led to the decision to order a pharmacogenetic test.

Developers of pharmacogenetic tools advocate for pre-emptive (i.e. prior to prescribing) use of their tools but we [10] have argued these tools are more likely to be ordered in a reactive (post-prescribing) fashion. A recent case series has shown the utility of pharmacogenetic testing for a patient that had failed to respond to multiple medications and for another that was experiencing a high side-effect burden [26], limitations of this evidence type notwithstanding, and that other mechanisms such as the nocebo effect may be operative [27]. As such, the evidence base for both pre-emptive and reactive use of pharmacogenetic tools for antidepressant therapy remains limited. Furthermore, there is debate around whether pharmacogenetic tools will boost antidepressant treatment adherence, with some arguing pharmacogenetic tools may assist in doctor-patient shared decision making, improve health literacy, reduce the perception of side effects and poor efficacy or treatment, and reduce health care costs [28]. In fact, a recent retrospective study of claims data implied an increase in adherence among patients for whom a polygene pharmacogenetic tool was ordered compared to those who received standard treatment [29]. Although prospective trials will be need to confirm these findings, promoting adherence is clearly desirable given data suggesting 42% of patients discontinue antidepressant treatment after 12 weeks [30] and long-term adherence is estimated to be 45% [31].

Three recent reviews have assessed the evidence-base of pharmacogenetic-guided antidepressant prescribing using different evaluation frameworks [10, 17, 32]. All reviews highlight that only a small proportion (<20%) of current pharmacogenetic tools have been empirically evaluated and suggest that pharmacogenetics has potential for clinical utility but there are numerous gaps in the current evidence. For example, it is unclear whether these pharmacogenetic-guided tools can shorten the time to remission and/or sustain the duration of remission from depression. Furthermore, the utility of these tools for assisting in the decision to switch or augment a patient’s current prescription remains uncertain, although preliminary evidence suggests these tools may assist with medication changes [33]. Finally, the cost-effectiveness of pharmacogenetic decision support tools remains unclear [34] and until robust economic studies are conduced, firm clinical guidelines or recommendations cannot be made.

There are a number of factors to consider if and when selecting a pharmacogenetic decision support tool and a variety of evaluation frameworks are available to clinicians to guide this process. One of the most well-known is the Oxford Centre for Evidence-Based Medicine (CEBM) Levels of Evidence [35]. The CEBM Levels of Evidence assists clinicians in identifying and appraising evidence using a hierarchy of the likely best evidence. In this hierarchy, systematic reviews of RCTs and individual RCTs are the preferred sources of evidence for making appraisals about a particular intervention. As mentioned above, three systematic reviews of these tools for antidepressant therapy have been completed [10, 17, 32] and two individual RCTs have been conducted [36, 37]. The CEBM advises that even in cases where supportive evidence exists, clinicians should consider at minimum four questions before choosing to use or adopt the intervention. Below we present these questions and provide salient information to assist clinicians in forming their own conclusions.

Ethnicity and its accompanying cultural and environmental factors account for some inter-individual pharmacokinetic and pharmacodynamic genetic variation relevant to antidepressant therapy [38]. For example, the alleles used to predict CYP2D6 and CYP2C19 metaboliser status vary considerably in frequency between ethnicities [39‐41] and tools including CYP2D6 and CYP2C19 do not necessarily measure the same alleles. This is a particularly important issue when contemplating whether to order a pharmacogenetic test for a non-Caucasian patient, in that most of the tools were developed and tested in Caucasian populations and may not include alleles that are rare in in this population but more frequent in people of Asian and/or African descent. As a result, non-Caucasians may be reported as normal (i.e. extensive) metabolisers by default when in fact they are poor or ultra-rapid metabolisers. Thus, it is important to be aware that ‘predicted’ metaboliser status provided by all pharmacogenetic tools should always be interpreted in the context of diverse cultural and environmental factors. In fact, a recent comparison of genotype-predicted and ‘true’ CYP2D6 metabolism showed genotype-predicted metaboliser status missed 43% to 64% of true poor metabolisers, depending on ethnicity [42]. It should also be noted that within the broad Caucasian, Asian, and African ethnic groupings, allelic differences in key pharmacogenes (e.g. CYP2D6 and CYP2C19) have been observed, although more subtle than difference observed between these broad groupings [43‐45]. Such ethnic and cultural variations need to be addressed in future pharmacogenetic tool development.

Another potential risk is loss of genetic privacy. Although privacy concerns are not unique to pharmacogenetic testing, it has been argued that genetic data is perceived as being higher quality and more definitive than other laboratory data, suggesting special protections are warranted [54]. However, not all genetic data is equal. Most antidepressant pharmacogenetic tools do not included genetic variants that are used to identify risk or diagnosis of disease and as such likely do not require additional privacy measures beyond those in place for other laboratory and clinical data. However, we are aware of pharmacogenetic tools that measure genetic variation in apolipoprotein E (APOE), a gene with potential risk implications for Alzheimer disease, as well as emerging tools that will employ genome and exome sequencing technology that have the capability of identifying disease-related mutations. Thus, mitigation of perceived and real genetic privacy concerns will continually need to be addressed by regulators, developers, and end-users of pharmacogenetic tools. To date, these issues have been addressed in some countries via genetic discrimination legislation such as the US Genetic Information Nondiscrimination Act of 2008 [55]. Most pharmacogenetic companies and laboratories offering pharmacogenetic testing use encrypted emails and password-protected websites to gain access to genetic information. However, the potential for this information to be inadvertently shared outside of the patient-clinician relationship is not trivial and for some patients could reduce their enthusiasm to be tested [56].