Use of a Consultation Service Following Pharmacogenetic Testing in Psychiatry

Abstract

The emerging discipline of pharmacogenetics (PGx) has the goal of aiding the selection of effective therapies and personalized dosing, decreasing the likelihood of adverse drug reactions and optimizing resource utilization. Simultaneously, the rapid evolution of economically feasible genetic testing technologies has resulted in a raft of commercial entities that provide genetic data to providers for use with their complex patients. The adoption of pharmacogenomics in psychiatry is growing, but it is limited by several factors, including the limitless permutations of drugs, comorbid conditions and concomitant medications and provider understanding of phenomena such as phenoconversion. We established an expert PGx consultation service for psychiatric providers who utilize our commercial PGx assay. To date, this service has provided ∼16,000 consults with extremely high levels of satisfaction; in an anonymous survey, 96% of respondents reported a rating of “very helpful” or “extremely helpful”.

Keywords: :

• adverse drug reactions
• consultation
• pharmacogenetics
• pharmacy
• psychiatry

The use of pharmacogenetic (PGx) testing in psychiatry is growing, supported by the need for better health treatments and emerging science supporting its clinical utility [1]. However, adoption is complicated by variable knowledge of PGx utility by providers and current lack of consensus within the field. Also complicating the use of PGx is the bewildering array of drug options for both psychiatric and other medical conditions. One such PGx assay lists 130 drugs used for psychiatry and/or chronic pain (Genomind, PA, USA). Because many psychiatry patients are on concomitant medications for comorbid conditions, this is further complicated by the more than 20,000 drug products representing over 2200 unique pharmaceutical ingredients approved for prescription use in the USA [2,3]. The majority of these drugs, in particular small molecule and non-biological chemical entities, are metabolized by hepatic enzymes, including the CYP450 superfamily of enzymes (phase 1 metabolism) and UDP glucoronyltransferases (UGT phase 2) [4]. These enzymes generally serve to convert active moieties to inactive metabolites that can be more easily eliminated, usually by increasing their water solubility. The routes of metabolism vary greatly, and complicating the matter is that some inactive moieties (i.e., prodrugs) require conversion to active metabolites via CYP450 enzymes. Variants of CYP2D6 and CYP2C19 are of particular importance for many psychotropic drugs [5].

The principle CYP450 enzymes responsible for phase 1 hepatic metabolism of pharmaceuticals include CYP1A2, 2B6, 2C19, 2C9, 2D6, 3A4 and 3A5. The genes encoding for these enzymes are polymorphic, with genes such as CYP2D6 having hundreds of different alleles associated with variable enzyme capacity [6,7]. The most common genotypes, which reflect standard metabolic activity, are classified as normal metabolizers (NM), formerly known as extensive metabolizers (EM). Other genotypes include poor metabolizers (PM), intermediate metabolizers (IM), rapid metabolizers (RM) or ultrarapid metabolizers (UM). Genetic variants characterized as PM can be found across CYP2B6, CYP2C9, CYP2C19 and CYP2D6 and encode for enzymes that are usually considered functionally inactive [7].

Factors including gender, age, nutritional status and comorbid conditions such as hepatic and renal insufficiency also influence the rate of metabolism (and elimination) of drugs. Perhaps most importantly, aside from their natural enzymatic capacity, CYP450 enzymes can be induced or inhibited by a dizzying array of exogenous agents including pharmaceuticals, herbal compounds, dietary compounds such as coffee and furanocoumarins (the latter found in grapefruit juice) and inhaled hydrocarbons (smoking). The conversion from an individual’s genotype-predicted capacity for metabolism to a different actual capacity is a phenomenon known as phenoconversion [6]. For example, consider an individual who is a NM at CYP2D6. We would expect normal metabolism of the beta-blocker metoprolol. This individual, however, is also taking bupropion, which is a potent CYP2D6 inhibitor. Bupropion converts this genetically normal metabolizer into a phenotypic poor metabolizer of metoprolol. In this scenario, metoprolol serum levels can be elevated three- to four-fold above those expected for a NM, representing a mismatch between the genotypically predicted exposure and the actual exposure, due to a drug–drug interaction.

In addition to the pharmacokinetic genes noted previously, pharmacodynamic genes also can affect drug efficacy and tolerability. Such genes typically encode for cell surface receptors, transporters, ion channel proteins, or other factors which are often the targets of drug action. Allelic variants of these pharmacodynamic genes have been associated with medication sensitivity or likelihood of response. For example, the HLA-B*15:02 risk allele is associated with dramatic increases in the likelihood of Stevens–Johnson Syndrome, a rare life-threatening drug-induced skin reaction following carbamazepine therapy [8].

According to data published by the National Center for Health Statistics, approximately 50% of individuals in the USA report taking at least one prescribed medication. Furthermore, 24% of individuals are prescribed three or more prescription medications and approximately 13% report the use of five or more prescription medications within the past 30 days [9]. In total 2.8 billion prescriptions were filled in 2000 [10]. Unfortunately, many patients will experience negative effects of medication usage, either in the form of adverse drug reactions (ADRs) or medication inefficacy [11]. ADRs are common, and these are associated with increased morbidity, mortality and resource utilization. Data published by the American Society of Pharmacovigilance identify ADRs that are linked to 1 million emergency room visits, 2.2 million hospital admissions, and approximately $136 billion in healthcare costs on an annual basis [12]. The true mortality burden of ADRs is unknown due to methodologic considerations, but it has been estimated that ADRs may be as high as the fourth leading cause of death [13].

Drug–drug interactions (DDIs) are one of the most common causes of ADRs. DDIs may result from direct interaction of co-administered drugs on drug targets, or from off-target toxicity, but more often are due to the propensity for many agents to induce or inhibit enzymes responsible for the metabolism of other, concomitantly administered drugs [14,15]. Previous estimates have cited that up to 22% of patients take interacting medications, with estimates up to 31% in the elderly population [16].

The US FDA requires drug manufacturers to conduct clinical pharmacology studies (known as phase 1 studies) to establish the absorption, distribution, metabolism and elimination routes of novel investigational agents, as well as to conduct drug–drug interaction studies under certain conditions. If approved for marketing, the resulting information is conveyed in the drug’s prescribing information or drug label. The FDA has PGx guidance in numerous approved drug labels, and for such agents, the Agency has summarized these drug warnings, interactions and precautions in a ‘Table of Pharmacogenomic Biomarkers in Drug Labeling’ [17]. In this table, which includes more than 270 drugs, psychotropics are the second most commonly listed therapeutic class of agents. In February of 2020, the FDA provided further PGx guidance with their ‘Table of Pharmacogenetic Associations’, which identifies additional medications which have sufficient scientific evidence of genetically influenced altered drug metabolism or adverse event risk but may not be contained in the drug’s prescribing information [18]. The FDA notes that “knowledge of a patient’s genotype may be used to aid in determining a therapeutic strategy, determining an appropriate dosage, or assessing the likelihood of benefit or toxicity”.

In addition to information provided in the drug labels, guidance is provided by the Clinical Pharmacogenetics Implementation Consortium (CPIC), an international consortium of academics, pharmacists and volunteers who provide independent guidelines based upon the available literature for gene–drug associations [19]. At this writing CPIC has published 26 such guidelines.

These factors create a daunting challenge for even the most expert prescribers because of the following reasons: A) no prescriber can be expected to remember all potential drug–drug interactions; B) the patient’s genotype is typically unknown; C) for some psychiatric drugs (with the exception of antipsychotics) the optimal therapeutic levels are undefined or poorly established; and D) for psychotropic medications, a clinical response is often not apparent for weeks [20].

The emerging science of PGx has the goal of aiding the selection of effective therapies and personalized dosing, decreasing the likelihood of adverse drug reactions and optimizing resource utilization. Simultaneously, the rapid evolution of economically feasible genetic testing technologies has resulted in the ability for many commercial entities to provide genetic data to healthcare providers for use with their complex patients.

There are a plethora of drugs available to psychiatrists for the treatment of mental illness, which can vary in efficacy, tolerability, metabolic pathways and drug–drug interactions. Many such psychiatric drugs are substrates, inducers, or inhibitors of the CYP450 enzymes enumerated above, and many non-psychiatric drugs are also capable of inducing or inhibiting CYP450 enzymes. Not surprisingly, patients’ response to these drugs varies widely. Our commercial laboratory, Genomind, Inc. (PA, USA), provides PGx testing that includes 24 pharmacokinetic and pharmacodynamic genes relevant for dozens of psychotropic drugs. In addition, we provide users access to gene–drug–drug interaction software that includes information on drugs accounting for >95% of all prescriptions written annually in the U.S.

Given the complexity of gene–drug interactions, the numerous possible gene–drug associations on the report, and the essentially limitless permutations of drugs, diagnoses and comorbidities, we provide clinicians with access to consultation by expert pharmacogeneticists as an aid in interpretation. Our database contains valuable information regarding the treatment of a diverse and challenging population.

Methods

Genomind^® Professional PGx Express^™ is a genetic assay utilized largely by psychiatric health care professionals. Using DNA obtained by buccal swab, the assay reports on 70 variants of 24 pharmacokinetic and pharmacodynamic genes, including ABCB1 (2 separate loci), CYP1A2, CYP2B6, CYP2C19, CYP2C9, CYP2D6, CYP3A4/5, UGT1A4, UGT2B15, SLC6A4, BDNF, MTHFR, ADRA2A, DRD2, MC4R, 5HT2C, ANK3, CACNA1C, OPRM1, GRIK1, HLA-A*31:01 and HLA-B*15:02. The report data are generated from Taqman^® real-time PCR assays using Thermo Fisher QuantStudio^™ 7 and 12 k platforms (Thermo Fisher Scientific, MA, USA). Results are collected from five principal assays including OpenArray^™, CYP2D6 copy number quantitation, and real-time PCR assays for SLC6A4 (SERT), HLA-A*31:01 and single SNP analysis. Earlier versions of this assay included 7, 10 and 18 genes [21,22].

Prescribers receive the PGx results in a dedicated HIPAA/HITECH compliant portal. With each assay, providers are made aware that they have the opportunity to obtain a telephone consult, which can be scheduled online. A prominent button is located on the provider portal to do so. Consultants are pharmacists (PharmDs) or PhDs with extensive training on the assay as well as PGx guidelines and best practices. Consultants are not clinical care providers but typically have additional training in genetics, neurohealth or psychiatry in addition to their graduate training. Consultations typically last up to 30 min.

Consultations conducted from 15 October 2018 through 31 August 2020 are included in this evaluation. The start date of 15 October 2018 coincides with the launch of a dedicated consult software upgrade which allows consultants to log clinical data into discrete fields. Standard operating procedures for consultations include the following:

1)	A review of the patient requisition form that includes diagnoses, failed and/or current medications. Any current medications are populated into a proprietary drug–drug–gene interaction software for assessment of phenoconversion risk and to review any published CPIC, DPWG or FDA PGx guidance/warnings. An assessment of potential alternative medications may be assessed also.
2)	After confirming patient name and date of birth, a brief review of current psychiatric symptoms is requested from the provider, including severity of symptoms via clinical global impression-severity (CGI-S) scale. Confirmation of current medication list occurs and is entered into software program. Any relevant PGx guidelines are discussed and any “strong” CYP450/p-glycoprotein or UGT inhibitors or inducers are highlighted. The definition of a strong inhibitor or inducer is taken from the FDA ‘Drug Development and Drug Interactions Table of Substrates, Inhibitors and Inducers’ [23].
3)	Consultations are conducted via videoconference (Zoom) with screen sharing features. A copy of the PGx report is screen-shared along with drug–drug–gene interaction software. Patients may participate in the consult only in the (actual or electronic) presence of their provider.
4)	Consultations are conducted in the context of standard treatment guidelines for the diagnosis/symptoms under consideration. Medications are usually discussed via a tiering system based on a published treatment algorithm. (i.e., SSRI/SNRI first line, TCA second line, MAOI third line for depression: CANMAT 2016 MDD Guideline [24]) and PGx gene–drug results are layered on top of these guidelines. Consultants have immediate access to the most common published treatment algorithms for major depressive disorder, generalized anxiety disorder, bipolar disease, attention deficit hyperactivity disorder, obsessive compulsive disorder, post-traumatic stress, and pain management. If a relevant drug–gene pair has a published PGx guideline, recommendations consistent with that guideline may be made. If a drug–gene pair does not have a guideline, as is the case with most pharmacodynamic drug–gene pairs, then a reference to core literature is made without a specific recommendation. For example, an individual with an MC4R gene associated with higher risk of antipsychotic induced weight gain would be apprised of the literature linking AA homozygotes to about a two-fold greater increase in weight [25]. While no specific recommendation would be given, consultants could help a prescriber sort second generation antipsychotics by inherent weight gain risk or provide additional literature regarding agents used to treat/prevent antipsychotic weight gain. In the case of pharmacokinetic gene–drug pairs that do not have a guideline, an estimate of the expected increase or decrease in AUC is provided to the clinician, without any specific dosing recommendation. For example, the antidepressant mirtazapine has no current PGx guideline, but a recent meta-analysis estimated a 39% to 51% increase in exposure for IM/PM compared to NM [26].
5)	During consultations, it is common for the discussion to stray outside of the scope of PGx and into general pharmacology. Consultants are trained to provide answers to these questions to the best of their ability or to follow up with a researched, written response if the question is outside of their expertise.

Following the consultation, consultants record information into discrete data fields, including the aforementioned CGI-S rating, medications previously or currently used, reason for prior drug failures and the pharmacogenetics of potential future medications.

At the end of the consultation, providers are asked to rate its value on a 5-point Likert scale, ranging from ‘not helpful’ to ‘extremely helpful’. Additionally, an online satisfaction survey was sent after each consult. Between October 2018 and August 2020, online satisfaction surveys (the validation surveys) were sent to 128 providers using an identical Likert scale as discussed earlier.

Results

From 15 October 15 2018 through 31 August 2020, a total of 94,910 PGx tests were completed among 8029 healthcare providers. Of these tests, 6401 were accompanied by a consultation (6.7%). The majority of these consultations were with US clinicians (6032), but a small portion included international providers (369). The US consultations were spread across 2205 prescribers for a mean of 2.73 consults per provider. 22% of these providers were using the consult service for the first time. Psychiatric providers are encouraged to have their patients attend the consults, and 457 patients joined the consultant and prescriber during this time period (7%).

The mean duration of a consultation was 21 min. Patient characteristics captured during consultations via a CGI-S scale revealed that the majority of patients (54%) were moderately ill, with the second largest proportion being markedly ill (23%); 4% were rated as ‘severely’ or ‘among the most extremely’ ill. The most frequent symptoms reported were depression, anxiety, insomnia and inattentiveness. The most common reason for prior drug failure was ‘inefficacy/inadequate efficacy’, followed by drug induced ‘agitation, irritability and/or anxiety’ and lastly ‘inadequate trial’. Selective serotonin reuptake inhibitors (SSRIs) were the most common class of discontinued drug; sertraline, escitalopram and fluoxetine were the three most commonly reported discontinuations and were also the three most likely to be associated with ‘no improvement’. Aripiprazole was the most commonly reported discontinued second generation antipsychotic.

Of genes reported on the assay, the five most commonly discussed were SLC6A4, MTHFR, CACNA1C, COMT and BDNF. The three most commonly discussed drugs (non-supplement or vitamin) were fluoxetine, lithium and duloxetine.

Ninety-four percent of consultations were rated as ‘extremely’ or ’very helpful’ by the provider at the time of consult. An independent, online validation survey of 128 providers confirmed these ratings, with 96% reporting a rating of ‘very helpful’ or ‘extremely helpful’. In addition, 94% reported that these consults were superior to PGx consults provided through other laboratories.

Discussion

The barriers to implementation of PGx as a standard of care include a lack of prescriber awareness/education, poor insurance coverage and integration into current workflow systems. While medical and pharmacy schools have already begun to expand PGx programs, it will be incumbent on providers of testing to provide additional educational resources. The results of this research support the strong appetite for PGx continuing education, with consult ratings exceeding 95% ‘very’/’extremely helpful’. This can be attributed to multiple learning avenues during the consultation service, which include a review of treatment algorithms (including treatment resistance options), a drug–drug interaction evaluation, references to relevant literature in real time, in addition to the PGx support.

While consultation ratings are important, it is also important to assess the impact of the consultations on prescribing patterns. All providers who experience a consult have access to the same drug–drug–gene interaction software to use independently following the consultation. Whenever a medication change is made, providers are asked, “Has this influenced your decision on drug selection or drug dosing for this patient?” Internal data shows that 85.6% answered ’yes’ to this question at the time of this publication. In addition, a previous survey of PGx providers indicated that the results influenced their medication decisions for 93% of patients [21].

Even though it is impossible to capture the individualized nature of consultations, we were able to summarize some interesting areas of discussion. For example, the most common genes discussed were all pharmacodynamic genes. This is likely due to clinician’s familiarity with the CYP450 system compared to pharmacodynamic genes such as SLC6A4, COMT or BDNF. For example, in a typical 20–30-min consult, it takes more time to introduce and explain the BDNF gene/protein and clinical studies than to dissect the significance of a CYP2D6 PM. In that light, these results are less surprising. Some may question the three most commonly discussed drugs as well…fluoxetine, lithium, duloxetine. When you consider that fluoxetine is the only SSRI without a PGx guideline and that duloxetine has no significant PGx interactions, then this becomes more palatable. Lithium is indicated for bipolar disease and commonly used for treatment resistant depression, which are two of the most common diagnoses for patients who receive our test. There are no strong PGx biomarkers for lithium either, so these three medications were more likely discussed as alternative treatment options to bypass potential PGx risk.

The large variety of psychotropic drugs available to providers, and their highly variable response rates, tolerability, capacity for drug–drug interactions and metabolic pathways presents a challenge for even expert psychopharmacologists. Consultation with experts in PGx provides additional useful information which may improve outcomes and decrease healthcare resource utilization. We are not aware of similar consult services for PGx in psychiatry.

Most providers who received consults saw high value in the consultation, as reflected by the fact that more consultations were provided to repeat-ordering clinicians compared to first-time users. Although 22% of new ordering clinicians used the service, it is not clear why a only a small fraction of repeat-ordering providers took advantage of the availability of consultations. The provider portal makes scheduling consults very convenient. It is likely that some providers do not take advantage due to time constraints, or that they do not see the need for such a consultation because they are sufficiently familiar with PGx, or that the initial consultation provided sufficient familiarity with the interpretation.

Conclusion & future perspective

A full review of the evidence supporting the utility of PGX is beyond the scope of this paper, but several lines of evidence suggest PGx testing in psychiatry and in general practice can improve outcomes and decrease resource utilization [1,27–29]. Perlis et al. found emergency room visits and hospitalizations were decreased by 40% and 58% respectively in the 6-month period following PGx testing, using the assay described in this report [22].

Limitations of these data include the fact that consultations were provided only for a minority of reports that were delivered to users. While users tended to see value in these consultations, real world outcomes data demonstrating this value have not yet been demonstrated.

This database may provide future opportunities for machine learning algorithms to further inform implications of included gene variants.

Executive summary

• Most drugs used in psychiatry are metabolized by CYP450 hepatic enzymes. The genes encoding these enzymes are polymorphic, and metabolic rates can vary greatly.

• Exogenous agents and polypharmacy contribute to the phenomenon of phenoconversion.

• Variants of genes encoding drug targets can also affect efficacy and tolerability.

• Drug–drug interactions are a common cause of adverse drug reactions.

• The use of pharmacogenetics is growing rapidly; its aim is to aid in the selection of safe and effective medications.

• A telephone consult service, staffed by expert pharmacogeneticists, was created to assist in the interpretation of a commercial pharmacogenetic assay.

•  Of consults, 96% were rated as ’extremely helpful’ or ‘very helpful’.

• Consultation with expert pharmacogeneticists holds the potential to help increase response rates and decrease adverse drug reactions and resource utilization.

Financial & competing interests disclosure

D Krause and D Dowd are employees of Genomind, Inc. and hold equity. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.