Clinical application of pharmacogenetics in pain management

Abstract

There is growing experience translating genomic data into clinical practice, as seen with the Implementing GeNomics In pracTicE (IGNITE) network. A primary example is the influence of CYP2D6 genotype on the beneficial and adverse effects of some opioids. Clinical recommendations exist to guide drug therapy based on CYP2D6 genotype for codeine, tramadol, oxycodone and hydrocodone, although the level of supporting evidence differs by drug. Limited evidence also supports the use of genetic data to guide other medications in chronic pain therapy, including tricyclic antidepressants and celecoxib. Pragmatic clinical trial data are needed in this area to better understand the impact of diverse populations, therapeutic interventions and clinical care environments on genotype-guided drug therapy for chronic pain.

Taking a personalized approach to patient care can allow clinicians to use individuals’ phenotype and genotype data (e.g., molecular profiling, medical imaging, lifestyle data) to identify the optimal treatment strategy, determine predisposition to disease and/or deliver targeted interventions [1]. Pharmacogenetics, or the use of genetic data to guide drug therapies, is a clinical tool within the personalized medicine toolbox [2]. Clinical evidence and resources that provide guidance to clinicians on the use of pharmacogenetic information for select gene–drug pairs are increasing [3]. These resources include pharmacogenetic data within the US FDA-labeling for nearly 200 medications and the availability of curated and ranked clinical pharmacogenetic evidence summaries in the Pharmacogenomics KnowledgeBase [4]. In addition, clinical guidelines that provide consensus recommendations on genotype-informed drug dosing are available for 35 drugs or drug classes from the Clinical Pharmacogenetics Implementation Consortium (CPIC), an international consortium developed in 2009 [5]. Additional clinical recommendations are provided for many gene–drug pairs from the Dutch Pharmacogenetics Working Group, the Canadian Pharmacogenomics Network for Drug Safety and the EMA [6,7].

Although clinical information is increasingly available, routine adoption of pharmacogenomics in clinical practice remains limited due to implementation barriers such as clinician inexperience with pharmacogenetic information and lack of integration of genomic results within the electronic health record with clinical decision support [8]. The University of Florida (UF) Health Personalized Medicine Program (FL, USA), a member of the NIH-funded Implementing GeNomics In pracTicE (IGNITE) network, was established to improve integration of genomic data into clinical practice. The UF Health Personalized Medicine Program launched in 2012 with CYP2C19 testing to guide antiplatelet therapy in patients undergoing percutaneous coronary intervention [9]. We have since expanded on this initial implementation to include, among other areas, the treatment of chronic pain [10]. The purpose of this report is to summarize selected gene–drug pairs with clinical relevance in pain management and present an example of how these pharmacogenetic data have been used in practice.

Chronic pain management

Approximately 11% of adults in the USA experience some form of chronic pain [11]. Symptom assessment and treatment recommendations for pain are often based on subjective data, which lead to challenges clinically. The WHO analgesic ladder provides an approach to managing patients with malignant pain. In practice, this ladder is often applied to nonmalignant chronic pain, although its application in this setting is limited by its original design for a malignant pain model and omission of pain etiology or biological mechanism in symptom assessment [12–15]. The WHO analgesic ladder directs therapy based on pain severity and persistence, with drug therapy recommendations spanning from nonopioids (e.g., NSAIDs) for mild pain followed by ‘weak’ (e.g., codeine) and ‘strong’ opioids (e.g., morphine) for moderate to severe pain, with inclusion of adjuvant therapies (e.g., duloxetine) as appropriate.

One in five patients seen by a clinician for pain receives a prescription for opioids in the USA [16]. However, a Cochrane analysis found long-term opioid therapy was either ineffective or poorly tolerated by a third of patients with nonmalignant chronic pain [17]. This lack of evidence for long-term therapy is also problematic for nonopioid therapy. Short-term opioid therapy has been associated with an adverse effect (e.g., nausea, constipation, somnolence, dizziness and pruritus) in 50–80% of patients [18,19]. An additional concern is possible addiction to prescribed opioids, although current pharmacogenetic evidence is unable to identify populations at increased addiction risk [20].

Pharmacogenetics & chronic pain management

A portion of the variability in patients’ response to pain medications can be attributed to genetic factors, including variation in genes encoding drug metabolizing enzymes (e.g., CYP450 enzymes). Knowledge of an individual’s pharmacogenetic data can provide an objective measure to help predict effectiveness and tolerability of medications in chronic pain management. However, it is essential for clinicians to recognize that although the effect of genetic variability on drug levels and patient outcomes have been studied extensively, the level of evidence supporting the clinical usefulness of genotype-guided drug therapy differs widely among gene–drug pairs. CPIC uses a systematic method to prioritize and assign evidence levels to gene–drug pairs, known as CPIC levels. CPIC levels are valuable tools that provide guidance for clinicians to interpret pharmacogenetic test results [21]. A CPIC level of ‘A’ or ‘B’ indicates that a gene–drug pair has ‘prescribing action recommended; alternative therapies or dosing are highly likely to be effective and safe’ [22].

To ensure a focus on current, clinically relevant pharmacogenetic information, this manuscript includes discussion of pain management gene–drug pairs that have a CPIC level of A or B (Table 2), including selected opioids, celecoxib, tricyclic antidepressants (TCAs) and serotonin and norepinephrine reuptake inhibitors (SNRIs) [21,22].

Table 2.

Clinical Pharmacogenetics Implementation Consortium level for opioids.

Medication	Drug class	Gene(s) involved	CPIC level†
Amitriptyline‡	TCA	CYP2C19	A
		CYP2D6	A
Celecoxib	NSAID	CYP2C9	B
Codeine	Weak opioid	CYP2D6	A
Methadone	Strong opioid	CYP2B6	B
Nortriptyline‡	TCA	CYP2D6	A
Oxycodone	Strong opioid	CYP2D6	A
Tramadol	Weak opioid SNRI	CYP2D6	A
Venlafaxine	SNRI	CYP2D6	B

^†Level A: genetic information should be used to change prescribing of affected drug; Level B: genetic information could be used to change prescribing of the affected drug because alternative therapies/dosing are extremely likely to be as effective and as safe as nongenetically based dosing. Level C or D: there are published studies at varying levels of evidence and no prescribing actions are recommended. Additional information can be found via [22].

^‡Evidence for other TCAs is based on amitriptyline and nortriptyline, therefore CPIC has assigned a Level B for CYP2D6 and CYP2C19 for tertiary amines (e.g., clomipramine) and a Level

B for CYP2D6 and secondary amines (e.g., desipramine) [24].

CPIC: Clinical Pharmacogenetics Implementation Consortium; SNRI: Serotonin–norepinephrine reuptake inhibitor; TCA: Tricyclic antidepressant.

CYP2D6 & opioids

CYP2D6 is drug-metabolizing enzyme encoded by a highly polymorphic gene. The CYP2D6 enzyme is involved in the metabolism of approximately 25% of clinically used medications, and genetic variations in the CYP2D6 gene are associated with changes in enzyme activity [25]. CYP2D6 genetic variability can lead to patients having one of four CYP2D6 phenotypes (i.e., poor, intermediate, normal and ultrarapid metabolizers). Approximately 5–10% of patients have complete deletion of the CYP2D6 gene (no CYP2D6 enzyme activity) and are classified as poor metabolizers. An estimated 1–2% of individuals have multiple copies of the CYP2D6 gene with increased enzyme activity [23,26]. The phenotype frequencies vary by race with certain race groups displaying higher (e.g., ultra-rapid metabolism in Middle Eastern populations) or lower (e.g., poor metabolism in east Asians) frequencies [27].

Codeine, tramadol, oxycodone and hydrocodone undergo bioactivation to more active forms via the CYP2D6 enzyme to varying extents. The clinical effects of genetic variability are especially important with codeine and tramadol, which are activated by CYP2D6 to morphine and O-desmethyltramadol, respectively [23,28]. These metabolites each have 200-fold greater affinity for the μ-opioid receptor than their parent compounds [29]. The CYP2D6-mediated metabolite is oxymorphone for oxycodone and hydromorphone for hydrocodone which have a 40-fold and 10- to 33-fold higher affinity for the μ-opioid receptor than their parent drug, respectively [23]. Patients who are CYP2D6 poor metabolizers have reduced concentrations of the active metabolites and are at risk for reduced analgesia with codeine, tramadol, oxycodone or hydrocodone [30–35]. Conversely, patients who are CYP2D6 ultra-rapid metabolizers have increased concentrations of the active metabolites and are at increased risk for toxicity [31–37]. There have been multiple cases of toxicity and/or death in pediatric patients receiving codeine post-tonsillectomy and/or adenoidectomy in which patients were discovered to be CYP2D6 ultra-rapid metabolizers postmortem [38,39]. The FDA added a boxed warning to codeine-containing products in light of these cases, which was subsequently updated to include a contraindication to codeine and tramadol use in patients <12 years of age in 2017 [40–42].

Although bioactivation of codeine, tramadol, oxycodone and hydrocodone is affected by CYP2D6 enzyme activity, the evidence supporting a link between CYP2D6 genotype and clinically relevant outcomes (e.g., pain relief ) differs by drug, with codeine having the highest level of evidence, followed by tramadol, with lesser evidence supporting clinically relevant effects with oxycodone and hydrocodone. This is explained to a great extent by pharmacokinetic differences among these agents. The bioactivation processes for oxycodone and hydrocodone include enzymes other than CYP2D6, providing alternative metabolic pathways. In addition, the parent oxycodone and hydrocodone compounds themselves have analgesic activity, allowing them to provide therapeutic effects in the absence of CYP2D6 activity [43,44]. For oxycodone, alternative pathways appear to be important as decreased CYP2D6 activity has been associated with decreased metabolite levels without changes in clinical outcomes [34].

CPIC guidelines provide CYP2D6 genotype-based recommendations for codeine (Table 1; CPIC level A), with additional recommendations included for tramadol, oxycodone and hydrocodone based on the strength of supporting clinical evidence [23]. Clinical recommendations based on CYP2D6 genetic variability are also provided by the Dutch Pharmacogenetics Working Group and the Canadian Pharmacogenomics Network for Drug Safety and are generally aligned with CPIC guidance [6,7].

Table 1.

Clinical Pharmacogenetic Implementation Consortium recommendations for medications by CYP2C19 and CYP2D6 phenotypes

Drugs	CYP2C19 phenotype	CYP2D6 phenotype	Recommendation
Codeine and tramadol	N/A	UM	Avoid due to increased risk of toxicity. Hydrocodone and oxycodone are also metabolized by CYP2D6 and are not good substitutes
		NM	Normal use of codeine and tramadol
		IM	Normal use but monitor for lack of response
		PM	Avoid due to risk of ineffective analgesia. Hydrocodone and oxycodone are also metabolized by CYP2D6 and are not good substitutes
TCAs – secondary amines (e.g., nortriptyline)	N/A	UM	Avoid use. Consider alternative drug not metabolized by CYP2D6
		NM	Standard therapy
		IM	Consider a 25% reduction of recommended starting dose and therapeutic drug monitoring to guide dose adjustments
		PM	Avoid use. Consider alternative drug not metabolized by CYP2D6
TCAs – tertiary amines (e.g., amitriptyline)	UM or RM	UM	Avoid use
		NM	Consider use of alternative drug not metabolized by CYP2C19
		IM	Consider use of alternative drug not metabolized by CYP2C19
		PM	Avoid use
	NM	UM	Avoid use. If use is warranted, consider titrating to a higher target dose
		NM	Standard therapy
		IM	Consider a 25% reduction in recommended starting dose
		PM	Avoid amitriptyline use and consider a 50% dose reduction
	IM	UM	Avoid use
		NM	Standard therapy
		IM	Consider a 25% reduction in recommended starting dose
		PM	Avoid amitriptyline use and consider a 50% dose reduction
	PM	UM	Avoid use
		NM	Avoid use. If use is warranted, consider a 50% reduction of recommended starting dose
		IM	Avoid use
		PM	Avoid use

Data taken from [23,24].

IM: Intermediate metabolizer; N/A: Not applicable; NM: Normal metabolizer; PM: Poor metabolizer; RM: Rapid metabolizer; TCA: Tricyclic antidepressant; UM: Ultra-rapid metabolizer.

CYP2D6 & CYP2C19 & antidepressants

The CYP2C19 enzyme, which is encoded by the CYP2C19 gene, and CYP2D6 are responsible for metabolizing many antidepressants, including drugs widely prescribed off label in the management of chronic pain (e.g., TCAs, SNRIs) [45]. CYP2C19 possesses genetic variation associated with a range of CYP2C19 enzyme activities and corresponding phenotypes (i.e., poor metabolism, intermediate metabolism, normal metabolism, rapid metabolism and ultra-rapid metabolism) [24].

Both CYP2D6 and CYP2C19 affect the metabolism of TCAs. Tertiary amine TCAs (e.g., amitriptyline) are metabolized by CYP2C19 to secondary amine TCAs (e.g., nortriptyline) which also possess pharmacological activity [24]. Secondary amines (e.g., nortriptyline) are then converted to less active metabolites via CYP2D6. Clinical outcomes have been associated with TCA serum concentrations and genotype (e.g., poor metabolism is associated with higher TCA concentrations and increased risk of adverse effects) [45–47]. CPIC guidelines exist to guide TCA dosing based on CYP2D6 and CYP2C19 genotypes (Table 1), although the association between TCAs and genotype may vary based on specific agent, dose and indication (Table 2; CPIC level A or B) [24].

Similar to TCAs, venlafaxine was primarily developed for use in depression but has since expanded to off-label use in chronic pain. Venlafaxine is an SNRI primarily metabolized by CYP2D6 to the active metabolite, O-desmethylvenlafaxine (ODV) [48,49]. Venlafaxine metabolizer status is often classified by either CYP2D6 genotype or the ODV/venlafaxine ratio [50,51]. One example of phenotype classification per ODV/venlafaxine ratio was a pooled analysis of four industry-sponsored studies that were initially used to gain FDA approval [50]. While there was no difference in dose between normal metabolizers and poor metabolizers, normal metabolizers were associated with higher ODV concentrations, and greater clinical improvement. The FDA package insert states CYP2D6 poor metabolizers have increased venlafaxine levels and reduced ODV levels compared with normal metabolizers but does not recommend dose adjustments based on CYP2D6 genotype [49]. The application of CYP2D6-guided venlafaxine therapy is complicated by factors including an additional active metabolite (i.e., desvenlafaxine), possible involvement of other hepatic enzymes (e.g., CYP2C19), and that the association with CYP2D6 may vary based on dose or indication (Table 2; CPIC level B).

CYP2C9 & celecoxib

CYP2C9 is responsible for the metabolic clearance of 15–20% of all drugs that undergo Phase I metabolism and is encoded by the highly polymorphic CYP2C9 gene [52]. FDA labeling for celecoxib indicates CYP2C9 is predominantly responsible for its metabolism and recommends initiating treatment at half of the lowest recommended dose in known or suspected CYP2C9 poor metabolizers [53]. Pharmacokinetic data have demonstrated that patients with decreased CYP2C9 activity have greater plasma concentrations of celecoxib, especially patients who are CYP2C9 poor metabolizers [54]. A randomized controlled trial of 282 children treated with a 3-day course of celecoxib after adenotonsillectomy found that patients who had a variant associated with a decreased function of the CYP2C9 enzyme reported less pain than patients without this variant. CPIC guidelines are not yet available for celecoxib (Table 2; CPIC level B).

CYP2B6 & methadone

CYP2B6 has high interindividual variation in CYP2B6 enzyme expression [55]. The CYP2B6 enzyme is involved in the metabolism of 4% of the top 200 medications, including methadone (Table 2; CPIC level B). Predicting drug response for methadone based on pharmacogenetic data is complicated by methadone’s complex pharmacokinetics and the number of enzymes involved in its metabolism. A systematic review and meta-analyses found that patients with decreased CYP2B6 activity had higher trough levels of the (R)- and (S)-enantiomers of methadone, more so with (S)-methadone [56]. However, these findings are based on limited data, with the availability of a small number of cross-sectional (n = 5) and case–control (n = 2) studies. Based on these factors the clinical application of CYP2B6 genotype in methadone therapy is debatable.

Translating pharmacogenetic data into practice: CYP2D6 codeine example

As the evidence supporting the role of genomic medicine mounts, the focus shifts to aid the development and investigation of genomic medicine practice models to enhance its implementation into routine clinical practice. Institutions affiliated with the IGNITE network and others have described such implementations [57–62]. Of the gene–drug pairs described in this report, clinical use of pharmacogenetic data to guide drug therapy is currently most prevalent with the CYP2D6 testing to inform codeine and/or tramadol prescribing. Varying clinical models have been proposed for implementation of pharmacogenetic testing generally [58–60,63], or for genotype-guided codeine therapy specifically [61]. When approaching implementation of CYP2D6 testing in practice to guide chronic pain management, some institutions have begun with a limited approach, such as targeting specific at-risk patient populations or limiting clinical recommendations to agents with the highest level of clinical evidence (i.e., codeine, tramadol). Gammal et al. implemented a pharmacogenetics-based prescribing strategy using clinical decision support to prevent the use of codeine in high-risk populations based on surgical history (i.e., after tonsillectomy, adenoidectomy) or high-risk CYP2D6 genotype (i.e., poor or ultra-rapid metabolism or indeterminate genotype) [61]. Their patient population consisted largely of patients with sickle cell disease in whom codeine was used extensively. Codeine was used in 173 of the 543 (32%) sickle cell disease patients without high-risk CYP2D6 genotype compared with 1 of 53 (4%) patients with a high-risk CYP2D6 genotype (the one patient with high-risk genotype who received codeine had previously tolerated codeine).

At UF Health, we are investigating a broader approach to use of pharmacogenetic testing through the Implementing Genomics in Practice (IGNITE) Proof of Concept Study: Genotyping in Family Medicine Clinics (ClinicalTrials.gov identifier NCT02335307). This ongoing pragmatic study compares pain management and pain control between patients with and without CYP2D6 genotype information available [64]. Patients with chronic pain, defined as pain for at least 3 months, are eligible for participation. The study utilizes a prospective cluster design in which primary care clinics are assigned as implementation or control sites. After providing written informed consent, a buccal cell sample is collected from patients at implementation sites, with genotyping completed in approximately 1–2 weeks. At that time, the physician caring for the patient receives genotype results and a pharmacist consultation that includes test interpretation and recommendations, with the final therapeutic decision being made by the physician. Control patients are offered genotyping, with results provided at study completion. The study is examining patient-reported pain outcomes, assessed using the NIH Patient Reported Outcomes Measurement Information System measures at baseline and 3 months. Change in patient-reported pain will be compared between patients enrolled from implementation versus control sites. Additionally, this trial will determine physician-perceived usefulness of CYP2D6 genotype results for prescribing decisions for chronic pain management.

Our experiences with this study to date have reinforced the importance of accounting for patient-specific factors that influence drug therapy selection, particularly in relation to medications that influence CYP2D6 activity (Table 3) [65,66]. This is especially important in patients who are CYP2D6 normal or intermediate metabolizers. In these patients, CYP2D6 inhibitors (e.g., duloxetine) may phenotypically ‘convert’ patients to CYP2D6 intermediate or poor metabolizers while the patient is taking the enzyme inhibitor. Second, several of the patients suspected to have abnormal CYP2D6 function by clinical criteria (e.g., poor responders to tramadol or codeine) were found to have normal a CYP2D6 genotype. This emphasizes the usefulness of genotyping to rule out an abnormal CYP2D6 genotype as a factor influencing drug response. Additionally, options for alternative therapy are limited in CYP2D6 poor and ultra-rapid metabolizers if a pain medication in the second step of the WHO analgesic ladder (e.g., codeine, tramadol, hydrocodone) is indicated. The primary alternatives for these patients require a step down (e.g., NSAIDs) or a step up (e.g., morphine), which may or may not be clinically appropriate [67]. For these patients, we have recommended an initial trial of oxycodone or hydrocodone with close monitoring, with step-up or step-down therapy to alternative medications as appropriate, since these opioids have the least evidence of clinically relevant effects caused by CYP2D6 variability [23,68]. Clinical experience with applying pharmacogenetic data in a pragmatic environment provides an important piece of the puzzle for clinicians desiring to use such tests in practice.

Table 3.

CYP2D6 inhibitors^†.

Strong inhibitors	Moderate inhibitors	Weak inhibitors
Bupropion, fluoxetine, paroxetine	Cimetidine, cinacalcet, duloxetine, fluvoxamine	Amiodarone, celecoxib, cimetidine, clobazam, cobicistat, desvenlafaxine, escitalopram, labetalol, ritonavir, sertraline

^†Strong CYP2D6 inhibitors reduce enzyme activity to the point of equivalence with poor metabolism while moderate inhibitors (e.g., duloxetine) reduce enzyme activity by approximately 50% [65].

Data taken from [69].

Conclusion

Pharmacogenetic-guided therapy for chronic pain is now a reality at some institutions for select medications (e.g., codeine, tramadol) metabolized by CYP2D6. In addition to pharmacogenetic data, providers should continue to consider other patient-specific factors (e.g., previous opioid use, renal function, concomitant medications) to guide drug therapy. As pharmacogenetics is translated from the research to the clinical setting, it is essential that the approach remains grounded in published evidence. The CPIC levels provide a clinician-friendly, systematic, evidenced-based tool to facilitate the application of gene–drug pairs appropriate for clinical care.

Future perspective

Many factors point to an expanded use of clinical pharmacogenetics. An increased number of medical centers have started offering genetic testing (including CYP2D6 testing) to predict drug response [57]. Regulatory bodies are beginning to more widely allow expanded use of genetic information as the FDA recently announced it will allow marketing of a direct-to-consumer genetic tests for disease risk [70]. In the future, clinical use of pharmacogenetic tests will likely become more prevalent as patient access to testing improves and the associated cost of testing decreases.

Evidence generation that incorporates pragmatic design and diverse populations may expedite a translation of pharmacogenetics into clinical care across a variety of settings, including nonacademic institutions. Pharmacogenetics is often referred to as a science of the outliers. That is, it is the patients at the extremes regarding drug effectiveness and response who may gain the most benefit from pharmacogenetic testing. As such, pharmacogenetic studies may be limited by sample size, as genotype is not expected to have a major influence on drug response for most patients. Alternative approaches to the traditional randomized control trial design are warranted to evaluate the impact of genotype-guided therapy on important patient outcomes given the expense and time required to conduct an adequately powered pharmacogenetic clinical trial. A pragmatic approach to evidence development may allow for examination of pharmacogenetics intervention in a real-world setting [71]. The UF Health Proof of Concept study described above is one example of a pragmatic trial design relying on practice-based evidence. Another example of an ongoing pragmatic study is the large, PREmptive Pharmacogenomic testing for prevention of Adverse drug REactions (PREPARE) study conducted by the Ubiquitous Pharmacogenomics Consortium in Europe [72]. PREPARE will enroll over 8000 patients in an attempt to quantify the collective clinical utility of implementing a panel of pharmacogenomic markers, including CYP2D6 testing for opioid response, into routine clinical care.

Executive summery

Chronic pain management

• Many patients and clinicians are faced with the difficulties of managing chronic pain.

• There is a need to improve evidence for long-term pharmacological pain management.

Pharmacogenetics & chronic pain management

• In clinical practice, pharmacogenetic tests for pain management are often ordered via commercial laboratories that use a panel-based approach that tests for multiple genes.

• The evidence varies for each gene–drug pair with select gene–drug pairs ready for implementation; every gene–drug pair is not created equally.

• Clinical Pharmacogenetics Implementation Consortium provides useful resources to guide evidence review and clinical implementation of pharmacogenetics.

• Recommendations are available for CYP2D6-guided therapy for codeine and tramadol; however, the evidence for hydrocodone and oxycodone are less strong.

• Of other available treatment options, CYP2D6 and CYP2C19 for tricyclic antidepressants, CYP2D6-venlafaxine, CYP2C9-celecoxib and CYP2B6-methadone have the best Clinical Pharmacogenetics Implementation Consortium level.

Translating pharmacogenetic data into practice: CYP2D6 codeine example

• Clinical decisions related to CYP2D6 genotype results must include patient factors that influence drug therapy selection, such as the patient’s history of response or toxicity, drug allergies or intolerances, and concomitant medications.

• Abnormal CYP2D6 enzyme function cannot be determined by clinical presentation alone; however, genotyping is able to rule out abnormal CYP2D6 genotype as a factor influencing drug response.

Conclusion

• Pharmacogenetic-guided therapy for chronic pain is a reality for select medications while other medications have developing evidence.

Future perspective

• We expect an increased use of clinical pharmacogenetics as costs and regulatory restrictions decrease in combination with improved evidence and access to testing