Pharmacogenetic Gene–Drug Associations in Pediatric Burn and Surgery Patients

Abstract

Management of critically ill patients requires simultaneous administration of many medications. Treatment for patient comorbidities may lead to drug–drug interactions which decrease drug efficacy or increase adverse reactions. Current practices rely on a one-size-fits-all dosing approach. Pharmacogenetic testing is generally reserved for addressing problems rather than used proactively to optimize care. We hypothesized that burn and surgery patients will have one or more genetic variants in drug metabolizing pathways used by one or more medications administered during the patient’s hospitalization. The aim of this study was to determine the frequency of variants with abnormal function in the primary drug pathways and identify which medications may be impacted. Genetic (19 whole exome and 11 whole genome) and medication data from 30 pediatric burn and surgery patients were analyzed to identify pharmacogene–drug associations. Nineteen patients were identified with predicted altered function in one or more of the following genes: CYP2C9, CYP2C19, CYP2D6, and CYP3A4. The majority had decreased function, except for several patients with CYP2C19 rapid or ultrarapid variants. Some drugs administered during hospitalization that rely on these pathways include hydrocodone, oxycodone, methadone, ibuprofen, ketorolac, celecoxib, diazepam, famotidine, diphenhydramine, and glycopyrrolate. Approximately one-third of the patients tested had functionally impactful genotypes in each of the primary drug metabolizing pathways. This study suggests that genetic variants may in part explain the vast variability in drug efficacy and suggests that future pharmacogenetics research may optimize dosing regimens.

Variability in drug efficacy is a major contributor to the difficulties in managing critically ill patients, especially in special populations (eg, burns, trauma, pediatric). Special populations may have more adverse consequences as a result of poor efficacy or adverse effects of drugs due to their altered metabolic states, a narrow therapeutic window and increased sensitivity to adverse effects. Pediatric burn and surgery patients are of particular interest given the larger number of medications administered simultaneously during their long hospitalizations. Pain medications, and particularly opioids, frequently demonstrate enormous variability in efficacy in these children, with resultant increased concerns regarding adverse effects. Opioids, such as fentanyl, hydromorphone, oxycodone, morphine, and methadone, are a mainstay for pain control in critically ill patients; however, the unpredictable response and toxic effects do not allow for a one-size-fits-all dosing approach, and there have been increasing concerns for other analgesics such as nonsteroidal anti-inflammatories (NSAIDs) and acetaminophen. Additional medications used to treat pediatric burn and surgery patients include but are not limited to serotonin 5-HT3 receptor antagonist, anticholinergic, N-methyl-d-aspartate receptor antagonist, corticosteroids, antimicrobials, antacids, and antiemetics.

Variability in drug efficacy is largely attributed to genetic variants in drug metabolism pathways. The cytochrome P450 enzymes (CYP) are the largest class of enzymes responsible for drug metabolism accounting for the primary route of metabolism for 70% to 80% of clinically used drugs.¹ These enzymes are in highest abundance in the liver but are also present in other organs such as the gastrointestinal tract, and biotransform endogenous and exogenous compounds to facilitate elimination. Most compounds are inactivated by CYPs, however, some, like morphine, are converted into active metabolites. The activity of CYPs may be altered during hypermetabolism, while coadministration of particular drugs can also cause induction or inhibition of CYPs’ activity.^2,3 Numerous CYP subtypes have been identified, with CYP2D6 and CYP3A4 being the most clinically significant.^4–6CYP2D6 and CYP3A4 are responsible for the metabolism of 25% and >50%, respectively, of medications in clinical use.⁶ Genetic variants associated with CYPs can result in reduced function. This can cause toxic accumulation of drug. Other variants have increased activity resulting in poor efficacy from low levels of drug in circulation. CYP2D6 polymorphisms occur more frequently and are associated with treatment failures and toxicity for several medications (eg, tamoxifen, propranolol, opioids).⁷ The reported allele frequency of CYP2D6 slow-metabolizing variants is reported to be disproportionate among different populations, with increased frequencies in Caucasians and Hispanic-Latinos, compared to African and Asian populations.⁸ Several other CYPs and genes (eg, CYP2B6, CYP2C9, CYP2C19, UGT1A, SULT1A1, transporters, receptors, and channels) are gaining more importance as knowledge regarding their clinical impact is discovered. The goal of pharmacogenetics is to study pharmacogene–drug associations to provide a molecular understanding of patient variability. To this end, genetic screening for not only CYPs but also other gene variants that have a high likelihood of influencing drug efficacy in patients would likely inform personalized dosing of medications.

Clinical decisions based in part on an individual’s genetic variants are a classic example of precision medicine. The majority of CYP clinical applications have involved cancer therapy.^9,10 For example, correlating therapeutic efficacy of tamoxifen in breast cancer patients with CYP2D6 activity has identified slow-metabolizers with greater risk for tamoxifen-induced toxicity, while fast-metabolizers exhibit increased risk for treatment failure and/or tumor recurrence. Another classic example is patients with human leukocyte antigen mutations developing Steven–Johnson syndrome or drug-induced hypersensitivity syndrome after certain medications are administered.¹¹ Only recently have studies suggested the use of CYP genotyping to improve dosing of common medications for pain management.^12,13 Pharmacogenetics is still a relatively young field and only in the past one to two decades has it really started to make better progress. Advances in genomic techniques, bioinformatics and a better understanding of molecular biology have allowed for the identification of associations between specific individual genetic variants and drug response. While much progress has been made to identify new genotype–phenotype associations by decoding the genetic variants of clinical importance in both metabolism and response pathways (such as the CYP2D6/Codeine relationship and the more recent discovery of CYP2C9 importance for celecoxib), there has not been widespread clinical implementation of pharmacogenetics guided prescribing. The reason for this lack of translation is multifactorial: 1) lack of knowledge on how to translate genetic information into clinical action, 2) interpretation and availability of genotype testing, and 3) lack of uniform recommendations for selecting drug/gene pairs. The Clinical Pharmacogenetics Implementation Consortium (CPIC) has attempted to fill these knowledge gaps. To date, 26 CPIC gene–drug guidelines have been published. Interestingly, four gene–drug guidelines are for the analgesics codeine (CYP2D6), tramadol (CYP2D6), oxycodone (CYP2D6), and celecoxib (CYP2C9).^14,15 To further advance the implementation of pharmacogenetics, and particularly incorporate pain management into precision medicine, we must conduct clinical studies in racially/ethnically diverse patient populations to provide robust data to inform clinicians and institutions of relevant and clinically important gene/drug combinations. This is of vital importance to provide individualized patient care, particularly related to pain management regimens, and decrease adverse outcomes.

As high-risk populations, pediatric severe burn patients and complex pediatric surgical patients require numerous medications during their hospitalization. In addition, they may require higher and more frequent opioid and analgesic doses, making them the ideal human model for pharmacogenetic research. Common medications include opioids (eg, fentanyl, hydromorphone), as well as compounds known to alter CYP activity (eg, propranolol, fluconazole).⁹ In a previous study in adult burn patients, we reported that, on average, a patient with >20% TBSA burn receives an average of 40.6 ± 20.2 medications per day.¹⁶ Unfortunately, drug dosing in burn patients does not truly account for pharmacokinetic, pharmacodynamic, and pharmacogenetic variation in this complex population. This may result in decreased efficacy, occurrence of adverse events, or development of tolerance and addiction.

Another important factor is that special populations may be at higher risk of phenoconversion, which is the phenomenon of a person’s genetic based phenotype being converted to a different category due to a nongenetic extrinsic factor.¹⁷ An example of phenoconversion is a normal metabolizer being converted to a poor metabolizer due to a disease state, such as children with sepsis being reported to have decreased CYP-mediated drug metabolism, thought to be related in part to an increased inflammatory state.¹⁸ Reduction in CYP activity attributed to inflammation has also been observed in patients following hip surgery, which demonstrated acute inflammation altered CYP activity in an isoform-specific manner.¹⁹ A thorough review by Lenoir et al also discussed numerous other inflammatory-related illnesses resulting in CYP inhibition including, infection, impaired organ function, diabetes, autoimmune disease, surgery, cancer, and administration of immune modulating medications.²⁰ Drug interactions also contribute to phenoconversion, such as antifungal and antiviral medications causing inhibition in CYP-mediated metabolism, while some antimicrobials and antiepileptics have been responsible for CYP induction resulting in increased metabolism.²¹ Moreover, hypermetabolism in burn patients results in severe alterations in organ blood flow, protein binding and synthesis, extreme inflammatory responses, as well as other comorbidities that can greatly impact drug metabolism.^22–24

Personalized dosing relies on a solid foundation of established pharmacokinetic/pharmacodynamic data, along with a thorough understanding of pharmacogenetics and drug–drug interactions. Thus, published data in burns and other special populations (eg, pediatrics) are insufficient for simulation modeling for projecting personalized dosing regimens in complicated patients. Thus, this study aims to elucidate the frequency of significant pharmacogene–drug associations in pediatric burn and surgery patients, specifically for medications they were administered during their hospitalization.

METHODS

Data utilized in this analysis were obtained from two ongoing studies to investigate fentanyl pharmacogenetics in pediatric patients: 1) samples collected in the intensive care unit (ICU) from pediatric acute burn patients and 2) samples collected in the operating room and post-op (OR) from pediatric surgery patients. Given that both groups are undergoing surgery and anesthesia, they are subjected to many of the same drugs and anesthetic treatment strategies. Both groups are being treated at the same institution in parallel with the same patient care staff and workflows. Hence, we have elected to combine data from the burn and surgical pediatric patients to conduct this pilot analysis. These studies were conducted under approved Institutional Review Board protocols reviewed by the WCG IRB.

Patient Enrollment

Thirty pediatric patients, ≤18 years old, were enrolled by direct informed consent and when necessary, guardian/parental consent. Inclusion criteria for the ICU study include: 1) TBSA burned ≥10%, 2) ability to consent or surrogate consent, and 3) receiving fentanyl therapy during dressing changes. Inclusion criteria for the OR study include: 1) having a surgical procedure, 2) ability to consent or surrogate consent, and 3) receiving fentanyl at the time of surgery. Exclusion criteria include: 1) nonsurvivable injuries and/or 2) condition where collection of research samples is unsafe (eg, severe anemia).

Sample and Data Collection

Blood specimens for genotyping were collected during the patient’s in-patient stay. The blood sample was drawn from an existing arterial or venous catheter line. Patient data were collected from their electronic medical records including demographics (eg, age, weight, ethnicity), all drug administrations, including doses and route, any concurrent identified infections, other medical conditions, as well as vital signs and routine laboratory results, and procedures (eg, surgery, line placement, etc.). All data were deidentified for analysis and patients were tracked by an assigned research study identifier. A list of all medications for each patient was extracted from the electronic medical records. Each list was curated to standardize medication names to generic names from brand names and to remove any duplications. Online databases (DrugBank, Pharmacogenetics Knowledgebase [PharmGKB], CPIC) and other published literature were used to determine primary and minor pathways for each medication.^25–30

Sequencing

Genomic DNA (gDNA) was isolated from whole blood using the Gentra Puregene Blood Kit (Qiagen, Maryland). DNA samples were assessed for quality and concentration. Isolated DNA samples were submitted to the University of California Sequencing Consortium laboratory for whole-exome sequencing (WES) (ICU Study) or to the Shriner’s Hospital for Children Genomics Institute for whole-genome sequencing (WGS) (OR Study) to achieve ≥×60 coverage. Raw sequence data (FASTQ) were mapped to the following human reference (GRCh37/hg19, December 2009) using the DRAGEN (Dynamic Read Analysis for Genomics) Ultra-Rapid Next Generation Sequencing Data Analysis Platform (Illumina).^31–33

Genomic Data Analysis

Full allelic decomposition, genotyping, and assignment of alleles for each pharmacogene were accomplished by applying the Aldy framework.^34,35 Aldy processes BAM alignment files to identify the genomic variants in the pharmacogenes, including single nucleotide variants (SNVs), insertions/deletions (Indels), copy number variants (CNVs), structural variants (SVs), and pseudogene fusions; star-allele discovery is guided by a database containing all known genomic variations associated with the different genotypes. Online databases, specifically the PharmGKB and Pharmacogene Variation Consortium (PharmVar) were used to assign a clinical function to each star-allele. Based on star-alleles identified, activity scores were assigned to each diplotype and patients were grouped into standardized phenotypic categories: poor, intermediate, normal (historically termed extensive metabolizer), rapid, or ultrarapid metabolizers, or unknown/indeterminate.^36,37

Statistical Analysis

Shapiro–Wilk test was used to evaluate continuous data (age, hospital length of stay, number of medications administered during hospitalization) for normality using GraphPad Prism v9.3.1 (GraphPad Software, California). Descriptive statistics was used to summarize cohort demographics, medication tallies, and length of hospitalization, with normally distributed data presented as mean ± standard deviation, while nonnormally distributed data are presented as median and range. Allele data are summarized by calculating frequencies within the studied population. As described above, the Aldy software utilizes star-allele databases to predict the most likely genotype for each patient dataset. Each Aldy solution has an associated confidence score with a maximum value of 1.0. By default, Aldy will only report out genotype results with a confidence score of 1.0.

RESULTS

A total of 30 patients were analyzed. Patients included 17 males and 13 females with a mean age of 10.3 ± 4.8 years. The patient ethnicities included 17 Hispanic-Latino (3 reported white, 14 other), 9 Caucasian (not Hispanic), 1 African-American, 1 Asian, and 2 unknown or declined to respond. The median length of stay for hospitalization was 27.5 days with a range of 151 days. Twenty of the patients were acute burn cases, 3 were previously burned patients returning for surgery and 10 were orthopedic spine surgery cases. Of the returning burn patients and orthopedic patients, the total length of stay was between 3 and 14 days. Preliminary analysis in 30 patients identified numerous functionally abnormal variants and several others of unknown clinical importance. Analysis was conducted using 19 WES and 11 WGS datasets and analyzed with the Aldy software to classify CYP star-alleles. Diplotype frequencies and predictions of functional impact are presented in Table 1. The wild-type (*1/*1) diplotype frequency was observed in approximately two-thirds of patients for both CYP2C9 and CYP2C19 and in just over half the patients for CYP2D6, whereas 90% of patients were wild-type for CYP3A4 but only 3.3% for CYP3A5. The remaining diplotypes have much lower frequencies apart from *3/*3 for CYP3A5 being represented in 80% of the patients. The CYP2D6 gene had the greatest number of different genotypes identified (n = 21), while CYP2C9 and CYP2C19 each had six different genotypes, CYP3A5 had four and CYP3A4 had three different genotypes. Nineteen patients were identified with predicted altered function in one or more of the following genes: CYP2C9, CYP2C19, CYP2D6, and CYP3A4 (Table 2). The majority had decreased function, except for several patients with CYP2C19 variants resulting in rapid (2) and ultrarapid (1) phenotypes. The CYP2C19*17 allele, which is responsible for the rapid and ultrarapid phenotypes, was only identified in WGS and not WES datasets. Numerous patients have alleles with unknown/indeterminate functional significance in CYP2D6 and CYP3A4. CYP2C9, CYP2C19, and CYP2D6 each have one-third of patients with an abnormal variant.

Table 1.

Diplotype activity scores and frequencies for CYP2C9, CYP2C19, CYP2D6, CYP3A4, and CYP3A5

Diplotypes in Study	Activity Score	Number of Subjects with Diplotype	Diplotype Frequency	Phenotype Prediction
CYP2C9
*1/*1	2	20	66.67	NM
*1/*2	1.5	6	20.00	IM
*1/*11	1.5	1	3.33	IM
*1/*3	1	1	3.33	IM
*2/*2	1	1	3.33	IM
*2/*3	0.5	1	3.33	PM
CYP2C19
*17/*17	4	1	3.33	UM
*1/*17	3	3	10.00	RM
*1/*1	2	19	63.33	NM
*2/*17	2	1	3.33	IM
*1/*2	1	5	16.67	IM
*1/*35	1	1	3.33	IM
CYP2D6
*1/*1	2	4	13.33	NM
*1/*2	2	3	10.00	NM
*2/*2	2	2	6.67	NM
*1/*35	2	1	3.33	NM
*2/*35	2	1	3.33	NM
*1/*41	1.5	1	3.33	NM
*2/*41	1.5	4	13.33	NM
*1/*4.021	1	1	3.33	IM
*2/*4	1	1	3.33	IM
*2/*4M+rs1135840	1	1	3.33	IM
*4C/*17	1	1	3.33	IM
*4/*35	1	1	3.33	IM
*4C/*35+rs28371703	1	1	3.33	IM
*41/*41	1	1	3.33	IM
*4.021/*41	0.5	1	3.33	IM
*4.021/*59	0.5	1	3.33	IM
*4+*4N.ALDY/*80	n/a	1	3.33	Indeterminate
*80/*83.ALDY or *61/*80	n/a	1	3.33	Indeterminate
*61/*79 or *79/*83.ALDY	n/a	1	3.33	Indeterminate
*1/*127	n/a	1	3.33	Indeterminate
*4.028/*139	n/a	1	3.33	Indeterminate
CYP3A4
*1/*1	n/a	27	90	NM
*1/*22	n/a	2	6.67	Indeterminate
*3/*22	n/a	1	3.33	Indeterminate
CYP3A5
*1/*1	n/a	1	3.33	NM
*1/*3	n/a	4	13.33	Indeterminate
*3/*3	0	24	80.00	Nonfunctional
*3/*7	0	1	3.33	Nonfunctional

IM, intermediate metabolizer; n/a, not applicable; NM, normal metabolizer; PM, poor metabolizer; RM, rapid metabolizer; UM, ultrarapid metabolizer.

Table 2.

Summary of abnormal phenotypes for CYP2C9, CYP2C19, CYP2D6, and CYP3A4

Patient	CYP2C9	CYP2C19	CYP2D6	CYP3A4	Patients With One or More Abnormal Function
1	IM	NM	NM	NM	A
2	NM	NM	NM	NM
3	NM	NM	IM	NM	A
4	IM	IM	IM	NM	A
5	NM	NM	NM	NM
6	PM	NM	UK	NM	A
7	NM	IM	UK	NM	A
8	IM	IM	IM	NM	A
9	NM	NM	NM	NM
10	NM	NM	NM	NM
11	IM	NM	NM	NM	A
12	NM	NM	NM	NM
13	IM	NM	NM	NM	A
14	NM	NM	NM	NM
15	NM	NM	NM	NM
16	IM	NM	IM	NM	A
17	NM	NM	IM	UK	A
18	NM	NM	NM	NM
19	NM	IM	IM	NM	A
20	NM	RM	NM	NM	A
21	NM	RM	NM	NM	A
22	NM	IM	NM	NM	A
23	NM	NM	NM	NM
24	IM	IM	IM	UK	A
25	NM	UM	IM	UK	A
26	IM	NM	NM	NM	A
27	NM	NM	UK	NM
28	NM	RM	UK	NM	A
29	IM	IM	IM	NM	A
30	NM	NM	UK	NM
Totals	10	11	9	3	19

A, abnormal phenotype; IM, intermediate metabolizer; NM, normal metabolizer; PM, poor metabolizer; RM, rapid metabolizer; UM, ultrarapid metabolizer.

Patients received an average of 43.9 ± 18.2 medications during their length of stay with a minimum of 19 and a maximum of 94. A total of 180 different medications were evaluated. Thirty-nine of the medications are reported to utilize CYPs for primary and/or minor routes of metabolism (Table 3). For the medications evaluated in this study, the CYP3A4 pathway is responsible for the largest number of medications for primary metabolism, followed by the CYP2D6 pathway, which had the greatest number of variants with predicted functional impact. Both CYP2C9 and CYP2C19 had approximately half the number of medications impacted as CYP2D6, however, had a similar number of patients, approximately a third, with variants with predicted functional impact (Table 2). The CYP3A5*3 allele was present in the majority of the patients and accounted for 53 of the 60 alleles analyzed. The CYP1A1, CYP1A2, CYP1E1, CYP2A6, CYP2B6, and CYP2C8 were identified in the medication metabolism review as playing a primary role in metabolism of 12 of the medications, however, variant evaluation to predict allele assignment and functional impact was not performed in this investigation.

Table 3.

Medications metabolized by CYPs and total anormal variants in primary and minor metabolism pathways

Medication	Totals for CYP2C9, 2C19, 2D6, and 3A4					Drug Class	Primary Metabolism CYPs	Minor Metabolism CYPs	Reference(s)
	P	P-UV	M	M-UV	N
Acetaminophen	NA	NA	8	7	13	Analgesic	2E1	1A2, 3A4, 2D6, 2A6	Krasniak et al (2014),27 DrugBank
Budesonide	0	0	NA	NA	0	Corticosteroid	3A4	3A5, 3A7	DrugBank
Buprenorphine	0	1			1	Partial opioid agonist	3A4		DrugBank
Celecoxib*	4	0	2	6	8	NSAID, COX-2 inhibitor	2C9	3A4, 2C8, 2D6	PharmGKB, DrugBank
Ciprofloxacin	NA	NA			NA	Second-generation fluoroquinolone	1A2		DrugBank
Clindamycin	0	0	NA	NA	0	Lincosamide	3A4	3A5	DrugBank
Clonidine	2	0			2	Alpha-2 adrenergic agonist	2D6, 1A2, 3A4, 3A5, 1A1		DrugBank
Dexamethasone	0	3	NA	NA	3	Corticosteroid–glucocorticoid	3A4	3A5	DrugBank
Dexmedetomidine	NA	NA			NA	Alpha 2 adrenergic agonist	2A6		Chidambaran et al38
Diazepam	5	2			5	Benzodiazepine	3A4, 2C19		PharmGKB, DrugBank
Diphenhydramine	4	3	14	0	12	Histamine H1-receptor antagonist	2D6	1A2, 2C9, 2C19	DrugBank
Duloxetine	1	0	1	0	1	Serotonin norepinephrine reuptake inhibitor	1A2, 2D6	2C9	DrugBank
Famotidine	4	0			4	H2 receptor antagonist	2C19		PharmGKB, DrugBank
Fentanyl	0	3			3	Opioid	3A4		PharmGKB, DrugBank
Glycopyrrolate	4	4	6	2	11	Muscarinic antagonist	2D6	1A2, 2B6, 2C9, 2C18, 2C19, 3A4	PharmGKB, DrugBank
Hydrocodone*	3	1			4	Opioid	2D6, CYP 3A4		PharmGKB, DrugBank
Hydrocortisone					0	Glucocorticoid	3A4, 3A5		DrugBank
Hydromorphone*			19	5	15	Opioid		2D6, 2C9, 3A4, 3A5	PharmGKB, DrugBank
Hydroxyzine			2	0	2	Histamine H1-receptor antagonist	3A4, 3A5		DrugBank
Ibuprofen*	8	0			6	NSAID, nonselective COX inhibitor	2C8, 2C9, 2C19		PharmGKB, DrugBank
Ketamine	0	1	12	0	12	NMDA antagonist	3A4	2B6, 2C9	Hijazi et al26
Ketorolac	3	0			3	NSAID, COX inhibitor	2C8, 2C9		DrugBank
Lidocaine	0	0			0	Local anesthetic	1A2, 3A4		P PharmGKB, DrugBank
Loratadine	2	1	3	0	3	Second-generation antihistamine	3A4, 2D6, 1A1, 2C19	1A2, 2B6, 2C8, 2C9, 3A5	DrugBank
Melatonin	NA	NA	6	0	5	Endogenous hormone	1A2	2C9, 2C19	Facciolá et al29
Methadone*	7	3	15	2	15	Opioid	3A4, 2B6, 2C19	2C9, 2C8, 2D6	DrugBank
Metoclopramide					0	Antiemetic, dopamine D2 antagonist	2D6, 3A4	1A2	DrugBank
Metoprolol	1	9			1	Beta-blocker	2D6	3A4	DrugBank
Midazolam	0	3			3	Benzodiazepine	3A4		PharmGKB, DrugBank
Minocycline			0	0	0	Tetracycline analog	3A4		Wehring et al28
Morphine*			1	0	1	Opioid		3A4	PharmGKB, DrugBank
Ondansetron*	0	3	8	4	12	Serotonin receptor antagonist	3A4	1A2, 2D6	PharmGKB, DrugBank
Oxycodone*	8	7	1	0	13	Opioid	3A4, 2D6		PharmGKB, DrugBank
Pantoprazole	2	0	0	0	2	Proton pump inhibitor	2C19	3A4	DrugBank
Propofol	9	0			9	GABA agonist	2B6, 2C9		PharmGKB, DrugBank
Risperidone	1	0			1	Second-generation antipsychotic	2D6		DrugBank
Ropivacaine	NA	NA	0	0	0	Local or regional anesthesia	1A2	3A4	Arlander et al (1998)39
Scopolamine	1	0	1	0	2	Belladonna alkaloid (anticholinergic)	3A4 suspected		DrugBank

COX, cyclooxygenase; CYP, cytochrome P450; GABA, gamma-aminobutyric acid; M, minor; N, total number of different patients with variants identified; NA, not analyzed; NMDA, N-methyl-d-aspartate receptor; NSAID, nonsteroidal anti-inflammatory drug; P, primary; UV, variant identified with unknown functional impact; X, variant identified in patient with predicted functional impact.

*Denotes the medications listed in CPIC guidelines related to CYPs.

The medications with the highest number of variants (seven or more) with predicted functional impact in the primary metabolism pathways included propofol (CYP2C9), oxycodone (CYP2D6), methadone (CYP2C19), and ibuprofen (CYP2C9 and CYP2C19). The medications with the greatest number of patients (11 or more) impacted in either the primary or minor metabolism pathways include hydromorphone, methadone, acetaminophen, oxycodone, ondansetron, ketamine, diphenhydramine, and glycopyrrolate.

DISCUSSION

Genetic variants impacting drug efficacy contribute to the interpatient variability that challenges patient care, yet the extent of this impact has yet to be fully understood. In this study, we have demonstrated the extensive array of variants present in significant drug metabolizing enzymes in a small cohort of patients. This is one of the first studies to report the pharmacogene–drug associations in pediatric burn and surgery patients. We found that these patients had a high frequency of genotypes with potential functional impact on clinically administered medications received during their hospitalization. Special populations, such as burn and surgery patients, may be at higher risk of experiencing the consequences of abnormal pharmacogene variants due to the large number of medications they receive not only during their hospitalization, but also after discharge. Thus, to improve pharmacologic prescription efficacy, we will need to evaluate patient genetic variants, identify the variant effects on medications, and identify potential drug–drug interactions due to genetic variants.

Analgesic and anesthetic drugs are among the most widely used medications in burn and surgery patients, as balanced anesthesia and pain management utilize a combination of medications from several drug classes. Approximately half of the drugs identified in this study that utilize CYP metabolism are analgesic and/or anesthetic compounds, with the medications being impacted the most by variants with functional changes all being in this category. This exemplifies the need for more focused research in pharmacogenetics specifically concentrating on analgesics/anesthetics in special populations to determine the functional impact in a clinical setting. As the initiative to drive precision medicine continues, pharmacogenetic studies are being published at exponential rates. Balyan et al recently reported significantly greater oxycodone exposure in pediatric surgical patients with poor and intermediate CYP2D6 phenotypes compared to normal and extensive metabolizers.⁴⁰ A study evaluating propofol in children found the drug distribution was higher in children with a UGT1A9 C allele and carriers of CYP2B6 T allele received significantly lower propofol doses, however, they could not conclude that CYP2C9 nor CYP2B6 polymorphisms could be used as independent predictors for propofol pharmacokinetics. This study concluded that further research needs to be conducted in these genes as well as other to explain the large interpatient variability observed with propofol.⁴¹ Additionally, several other studies have recently been published in children related to analgesic pharmacogenetics, including acetaminophen,^27,38 morphine,⁴² and fentanyl.⁴³ However, each of these studies only investigated one drug and focused on a small number of genes. Cohn et al conducted a broader study investigating pharmacogenetic implementation in pediatrics in a tertiary care setting and reported the drug–gene interactions associated with pain medications were identified in the opioids codeine, hydrocodone, oxycodone, and tramadol related to CYP2D6 and the NSAIDs celecoxib, flurbiprofen, ibuprofen, meloxicam, and piroxicam related to CYP2C9.⁴⁴ These findings are consistent with many of the medications and variants identified in our patient population.

CYP2C9

Our study demonstrated that one-third of the patients had abnormal alleles (*2, *3, *11) in CYP2C9, resulting in nine patients with intermediate and one patient with poor metabolizer phenotypes. The CYP2C9 pathway is estimated to biotransform approximately 15% of CYP-related drug metabolism and is responsible for the primary metabolism of NSAIDs such as celecoxib, ibuprofen, ketorolac, as well as propofol, voriconazole, and warfarin.^45,46 There is a CPIC guideline specifically for NSAIDs recommending a decreased dose for poor metabolizers, due to the risk of increased adverse effects such as gastrointestinal bleeding, with reduced elimination.¹⁵ There are currently 75 alleles reported in the PharmVar database. The most common variants investigated for CYP2C9 include the *2 and *3 alleles, which are reported to have decreased and null function, respectively.^1,45 Their frequencies are believed to be low in African and Asian populations, however, the *2 allele has been reported to be in approximately 10% to 16% of Caucasian populations, while the *3 allele is reported to be in 3% to 8% of Caucasian, Hispanic, and East Asian populations.^1,47 The CYP2C9*11 allele also has reduced function, has been reported in less than 1% of the population, has been identified in all ethnicities, and been associated with impaired warfarin metabolism.⁴⁵

CYP2C19

A third of our patients also had abnormal phenotypes in CYP2C19 as a result of the *2, *35, and *17 alleles. CYP2C19 is responsible for the metabolism of multiple analgesics, proton pump inhibitors, and other drugs including diazepam, famotidine, ibuprofen, loratadine, methadone, pantoprazole, and voriconazole.^44,48,49 There are two CPIC guidelines related to gene–drug associations identified in our study for proton pump inhibitors⁵⁰ and voriconazole.⁵¹ There are currently 36 alleles listed in the PharmVar database. The *2 and *35 alleles are both reported to be null in function, while the *17 allele has increased function. The *2 allele is the most common loss of function allele with frequencies approximately 15% in Caucasians and Africans, and 29% to 35% in Asians.⁵² The *35 allele has been reported primarily in African populations at a frequency of 1% to 3%. Interesting the one individual heterozygous for this allele was the only African-American patient enrolled in our study. The *17 allele is the only variant reported in CYP2C19 to have increased function, which is the result of a C > T transition in the promotor region resulting in increased expression and activity. The presence of *17 has been observed throughout many ethnic groups with frequencies ranging from 3% to 21%.⁵² In the current study, the *17 allele was only identified in WGS datasets. While WES is one of the most common approaches used in genomic studies and is much more cost effective compared to WGS, thus allowing a greater number of patients to be included in studies, it does not capture important variants in nonexonic regions, which is the case for CYP2C19*17.⁵³ Thus, it is highly likely we missed identifying numerous CYP2C19 rapid and ultrarapid metabolizers in our patients analyzed using WES.

CYP2D6

We have also identified several abnormal alleles in the CYP2D6 gene in this study. It is well known that the CYP2D6 gene is the most polymorphic pharmacogene. In the current study, the drugs most affected by variants in CYP2D6 include clonidine, diphenhydramine, duloxetine, glycopyrrolate, hydrocodone, loratadine, metoprolol, oxycodone, and risperidone. Related to the drugs identified in this study, there are two CPIC guidelines discussing several opioids⁵⁴ and ondansetron,¹⁴ however, guidance is limited due to a lack of definitive research. The PharmVar database currently lists 142 different alleles for CYP2D6. CYP2D6 is responsible for metabolizing approximately 25% of clinically used drugs, which includes analgesics (eg, opioids), antidepressants, antihistamines, and beta blockers.⁴⁹ Translation of CYP2D6 genotypes into phenotypic categories is complex and has been challenging due to the lack of standardization across clinical and research laboratories, resulting in discordant phenotype assignments. A recent publication from CPIC and the Dutch Pharmacogenetics Working group has addressed this issue recruiting several international experts to establish a standardized method for translating CYP2D6 genotypes to metabolizer phenotypes.³⁷ These refined activity score categories have been used in our study, along with defining wild-type equivalent variants as normal metabolizers rather than the historical term of extensive metabolizers.

The CYP2D6*2 allele has slightly reduced function and ranges from approximately 12% to 28% in most populations. The *41 allele also has slightly reduced function and is present in approximately 2% to 15% of the population. While *2 and *41 are reported to have slightly decreased function, the overall clinical impact does not appear significant and thus both are assigned as normal function, along with *35, which was also identified in our study. The *4 allele is nonfunctional and is the most common cause of poor metabolizers. It has been reported to be in approximately 19% of European populations, 12% of Latinos, less than 5% in African populations, and only about 1% of Asians. In the current study, the *4 allele was responsible for the majority of the intermediate metabolizer phenotypes. The *17 allele is reported to have reduced function and is typically observed in African populations with frequencies of 17% to 19%, while most other populations range from less than 1% to 3%. CYP2D6*17 was identified in the only African-American patient in our study cohort. While the *59 allele is reported to have normal function, when in combination with the *4 allele the resulting phenotype is an intermediate metabolizer. Numerous other CYP2D6 alleles were identified in our patient population that lack known functional impact. Reported population frequencies of some of the other alleles presented in our study can be found on the PharmGKB CYP2D6 frequency table (https://api.pharmgkb.org/v1/download/file/attachment/CYP2D6_frequency_table.xlsx).

CYP3A4/CYP3A5

Our study identified the CYP3A4 *3 allele in one patient and *22 allele in three patients. CYP3A4 and CYP3A5 are the predominate cytochrome P450s expressed in the liver. It has been reported that CYP3A4 is predominant in Caucasians, while CYP3A5 is predominant in Africans. CYP3A4 is responsible for 50% to 60% of clinical drug metabolism,^55,56 spanning a large range of drug classes, including but not limited to the opioids fentanyl, hydrocodone, methadone, and oxycodone, anesthetics clonidine, diazepam, midazolam, and ketamine, and antimicrobials clindamycin, minocycline, and voriconazole. There are currently 35 different CYP3A4 alleles listed in the PharmVar database, however, none to date have been assigned a clinical function. The *3 allele has been reported to have a frequency of approximately 1% in Caucasian populations.⁵⁶CYP3A4*22 has been reported in the literature to have reduced function and has been reported to have a 3% to 5% allele frequency predominately in Europeans and admixed Americans, and has affected the metabolism of midazolam, tacrolimus, cyclosporin, and statins.^55,57 All three patients in our study with the *3 and *22 alleles are Caucasian. Our study identified the CYP3A5 *1, *3, and *7 alleles. The *1 allele is reported to have normal function and both *3 and *7 are null in function. CYP3A5 has nine alleles reported in the PharmVar database with predicted functions for many. The PharmGKB CYP3A5 allele frequency table reports the *1 allele to be in approximately 45% of Africans, 25% to 33% of Asians, 17% of Latinos, and only 7% in European populations. The *3 allele is reported in approximately 92% of Europeans, 77% Latinos, 67% to 75% Asians, and 32% of Africans. Interestingly, the *7 has reported frequencies of 12% in Africans and 2% in Latinos but is not reported in Europeans and Asians. The majority of the *1 alleles identified in our study were in Latino patients with the one Asian patient being heterozygous for *1/*3. Our patient with the *3/*7 genotype was the only African-American in our study.

The present study is limited to a small number of patients included in this analysis along with decreased ethnic diversity. The majority of our patients are Hispanic-Latino and Caucasian, thus the genotype frequencies presented here are not a good representation of the general population. Larger cohorts with greater ethnic diversity, particularly African and Asian, are needed to truly capture the presence of rarer polymorphism. These rarer polymorphisms are also more likely to be classified as “indeterminate” function as we have presented several examples of for CYP2D6, CYP3A4, and CYP3A5. While there is limited research that is suggestive that some of these alleles have decreased function, eg, CYP3A4*22 or any of the CYP2D6 diplotypes that include *4, until there is a greater amount of evidence to support the clinical function, databases will be reluctant to change their classifications. There is a tremendous need to conduct clinical research in racially diverse populations, particularly in minorities, that not only characterizes the frequency of polymorphisms but also collects data to infer the functional impact of these variants. Future research also needs to focus on the impact of illness, comorbidities, and medical procedures on phenoconversion. Investigating the function of CYPs in patients is challenging without invasive techniques such as tissue biopsies, however, newer approaches such as exploring the utility of circulating biomarkers, eg, specific drugs that serve as biomarkers for a singular pathway or other approaches such as exosomes, may hold promise to identify clinical diagnostic approaches to guide precision dosing adjustments in individual patients.

Moreover, as stated above, another limitation is the WES datasets may have not captured the CYP2C19*17 allele, which is in a promoter region, in addition to other variants that are in nonexonic regions as well as duplicate genes. This limitation should be taken into great consideration for future pharmacogenetic study design. It is understandable that WES may be attractive to use due to its decreased cost, thus allowing more patients in a study, however the fact we are missing vital information for these patients is a disservice to the advancement of pharmacogenetic research and will hinder the progress of improving outcomes in patient management.

CONCLUSION

The high prevalence of functionally abnormal variants in combination with the high number of impacted medications administered to patients in this study demonstrated significant pharmacogene–drug associations in pediatric burn and surgery patients. This study provides robust clinical data that can be used for targeted future clinical pharmacogenomic studies, particularly in special populations.

Funding: These studies were funded by a Shriner’s Hospital for Children (Shriner’s) Developmental grant (79160 NCA-17-E), a Shriner’s Clinical Research grant (71013-NCA-20), and a National Institute of Health (NIH) Office of Research Infrastructure Programs K01 Career Development Award (1K01OD026608-01) and NIH Clinical Research Loan Repayment Program Award. The Genomics Shared Resource is supported by the University of California, Davis Comprehensive Cancer Center Support Grant (CCSG) awarded by the National Cancer Institute (NCI P30CA093373).

Conflict of interest statement. The authors have no conflicts of interest to declare.