Evidence for Pharmacogenomic Effects on Risperidone Outcomes in Pediatrics

Abstract

Objective:

To determine the association between genetic variants reported to impact risperidone and adverse events (AEs) in children and adolescents.

Methods:

Individuals ≤18 years with ≥4 weeks of risperidone exposure in a de-identified DNA biobank were included. The primary outcome was AE frequency as a function of genotype. Individuals were classified according to metabolizer status for CYP2D6, CYP3A4, and CYP3A5; wild type, heterozygote, or homozygote for specific single nucleotide variants for DRD2, DRD3, HTR2A, and HTR2C; and wild type versus non wild type for multiple uncommon variants in ABCG2, ABCB1, and HTR2C. Tests of association of each classification to AEs were performed using a Fisher’s exact test and logistic regression, and statistically significant classifications were included in a final logistic regression.

Results:

The final cohort included 257 individuals. AEs were more common in CYP2D6 poor/intermediate metabolizers (PMs/IMs) than normal/rapid/ultrarapid metabolizers (NMs/RMs/UMs) in univariate and multivariate analysis. HTR2A-rs6311 heterozygotes and homozygotes had fewer AEs than wild types in logistic regression but not in univariate analysis. In the final multivariable model adjusting for age, race, sex, and risperidone dose, AEs were associated with CYP2D6 (AOR 2.6, 95% CI 1.1–5.5, for PMs/IMs vs. NMs/RMs/UMs and HTR2A-rs6311 (AOR 0.6, 95% CI 0.4–0.9, for each variant allele), both consistent with prior studies.

Conclusion:

Children and adolescents who are CYP2D6 PMs/IMs may have increased risk for risperidone AEs. Of the genes and variants studied, only CYP2D6 has consistent association and sufficient data for clinical use while HTR2A-rs6311 has limited data and requires further study.

INTRODUCTION

Risperidone (Risperdal) is an antipsychotic that acts as a serotonin-dopamine antagonist. It has FDA approval for pediatric indications including bipolar disorder 1 ages 13–17, irritability in autism ages 5–27, and schizophrenia in children 10–17.¹ It is commonly prescribed off label in pediatrics for indications including attention deficit hyperactivity disorder (ADHD), Tourette syndrome, behavioral disturbances, and impulse control disorders.¹

Up to 1 in 3 children experience an adverse event (AE) while taking risperidone.² Risperidone AEs are more established in adults with the most common including metabolic abnormalities and weight gain.² There is less established data on pediatric risperidone AEs, although children are at higher risperidone AE risk compared to adults. There is some suggestion that children are more likely than adults to experience weight gain and metabolic abnormalities.³

Risperidone acts on both serotonin (5-HT_2A) and dopamine (D₂) receptors as an antagonist but with higher affinity for 5-HT_2A receptors.⁴ Multiple genes have been suggested to contribute to the safety and tolerability of risperidone.⁵ Metabolism of risperidone is primarily via hepatic cytochrome P450 (CYP) 2D6, while other enzymes, such as CYP3A4 and CYP3A5, are known to contribute to risperidone metabolism to a lesser extent.⁶ The ABCB1 transporter affects the efflux of a wide variety of drugs including risperidone. Variants in the genes encoding ABCB1 and ABCG2, another transporter protein, have been previously associated with risperidone drug concentrations or toxicity.⁷ There are many genes and variants that are known to affect risperidone response and toxicity through altering the pharmacodynamic targets and pharmacokinetic metabolic pathways of risperidone (Table, Supplemental Digital Content 1).⁵ These genes are often included on pharmacogenomic panels, and thus it is expected that clinicians will interpret these results and apply them to clinical drug prescribing as appropriate.

Pharmacogenomic testing is becoming increasingly available and more widely used by clinicians and patients to help guide prescription of drugs with high potential risks for AEs.⁸ These testing results often list the metabolizer class for a patient for each cytochrome P450 enzyme (such as CYP2C19 or CYP2D6). Each of these enzymes may have a spectrum of activity from no activity (in poor metabolizers) to normal activity (in normal metabolizers) to increased activity (in ultrarapid metabolizers).⁹ For various genes that are targets of psychiatric medications, such as serotonin and dopamine receptors, the reports may include specific single nucleotide polymorphisms (SNPs), identified by their reference SNP identifier (rsID) and reported as wild type, heterozygous, or homozygous.⁵

Clinical Pharmacogenomic Implementation Consortium (CPIC) or Royal Dutch Pharmacists Association Pharmacogenetics Working Group (DPWG) consensus guidelines report evidence based dosing guidance for drug-gene pairs. There are currently no society guidelines for risperidone dosing based on pharmacogenomics.^10,11 Various pharmacogenomic laboratory reports have included risperidone-prescribing guidance based on results for CYP2D6, CYP3A4, CYP3A5, HTR2A, HTR2C, DRD2, and DRD3. However, the body of evidence suggesting the clinical utility of the various genes commonly reported on these panels is mixed with few data for the pediatric population.^8,11 These prior studies are small in size, with fewer than 120 children, and often report conclusions that are not uniformly replicated in other studies.¹¹ There are also some genes that have been associated with risperidone toxicity in pediatric populations including ABCB1 and ABCG2 that are generally not tested on commercially available pharmacogenomic panels.⁷

In order to provide insight into the impact of previously reported genetic variants on pediatric risperidone safety, this retrospective cohort study assessed the association between genetic variants in several genes (ABCB1, ABCG2, CYP2D6, CYP3A4, CYP3A5, DRD2, DRD3, HTR2A, and HTR2C) and the risk for risperidone AEs in a large cohort of pediatric patients exposed to risperidone for at least 4 weeks.

METHODS

Study Design and Cohort

Data for this study were obtained from BioVU, Vanderbilt University Medical Center’s de-identified biobank linking DNA to electronic health records (EHR).^12–14 This study was reviewed by the Institutional Review Board and determined to be non-human subjects research. Inclusion criteria included children exposed to risperidone for ≥4 weeks; age ≤18 years at the start of risperidone; and a DNA sample that was non-compromised or unaffected by disease related changes. Exclusion criteria included medication management by a provider outside of the institution and inadequate follow up data, such as unclear medication dose or adverse event (AE) status. Specific examples of exclusion criteria were lack of notes indicating medical decision making or management of risperidone and lack of documentation regarding the presence or absence of AEs. Individuals who had ambiguous CYP2D6 metabolic status after DNA sequencing were also excluded from analysis. Data were collected for the individuals in the cohort from the time the medication was started until discontinuation of the medication or end of the study period (Dec 2017).

Primary Outcome and Identification

The primary outcome assessed in this study was the presence of AEs in children and adolescents during risperidone therapy, which were documented and attributed to risperidone in the EHR. As defined by the WHO, any untoward event or laboratory abnormality reported by the patient, guardian, or identified by clinicians was considered an AE.¹⁵ If AEs were documented, specific details surrounding the AE were noted including the specific type of AE, date of AE, dose of risperidone, and management by the prescriber. Date of AE was counted if specifically mentioned in the note or encounter, otherwise the date of the first note mentioning the AE was used as the date of AE. Since this was a retrospective study, no causality assessment was performed to establish the relationship between AEs and risperidone. One person manually reviewed each chart to determine AE status. This chart review was blinded to all genotypes in order to prevent reviewer bias when determining AE status.

Data Abstraction

Data for this study were collected and stored to REDCap. Each individual’s chart was reviewed for the following data: demographic information (sex, race, ethnicity, and age on risperidone initiation), clinical data (risperidone indication, mental health diagnoses, and other medical comorbidities), and medication information (risperidone dose, daily dosing schedule, duration of medication use, and concomitant drugs).

DNA Analysis

Each individual’s DNA was analyzed to determine metabolizer status for the cytochrome P450 genes (CYP2D6, CYP3A4, and CYP3A5), wild type, heterozygous for variant, or homozygous for variant for common single nucleotide polymorphisms of other genes (DRD2 (rs1799978 and rs6277), DRD3 (rs6280), HTR2A (rs6311), and HTR2C (rs1414334)), and presence or absence of at least one of the uncommon variants in ABCG2 (rs2231142), HTR2C (using rs3813929 and rs6318), and ABCB1 (using rs1045642, rs1128503, and rs2032582). DNA sequencing was performed using the Kailos TargetRich™ PGx Panel Next Generation sequencing assay (Kailos Genetics, Inc, Huntsville, AL) and was performed by the institutional DNA analysis core laboratory using reagents and protocols as specified by the manufacturer. Long-range PCR was used to determine CYP2D6 deletions or duplications. Genotypes for the cytochrome P450 enzymes were assigned based on CPIC guidelines, and then individuals were characterized as poor, intermediate, normal, rapid, or ultrarapid metabolizers.^9,16–18 Full details of the CYP2D6 variants included and assignment of functional status have been previously reported.²

Statistical Analysis

For statistical analysis, individuals were classified as wild type, heterozygotes, or homozygotes for the single nucleotide variants in DRD2, DRD3, HTR2A, and HTR2C. For CYP2D6, CYP3A4, and CYP3A5, individuals were grouped as poor/intermediate metabolizers (PMs/IMs) or normal/rapid/ultrarapid metabolizers (NMs/RMs/UMs) in order to utilize all genotyped individuals, and ensure that each comparator group had 10 or more individuals. For HTR2C, ABCB1, and ABCG2, individuals were classified as either wild type (no variant present) or non-wild type (at least one variant present) in order to ensure that each comparator group had 10 or more individuals. Data such as demographics, clinical variables, and outcomes were calculated using frequencies for categorical variables and medians with interquartile ranges (IQR) for continuous variables. Each gene was tested for association to the outcome (AEs) and the clinical characteristics using Fisher’s exact or Kruskal-Wallis test, as appropriate. Multivariable logistic regression was performed initially for each variant adjusting for dose, age, gender, and race. A final multivariable logistic regression was performed including any gene/variant that showed significance in initial multivariable analysis, adjusting for dose, age, gender, and race with outcome of AEs. STATA v15.1 (StataCorp, College Station, TX) was used for data analysis. Statistical tests were two sided and a P value <0.05 was considered statistically significant.

Potential covariates to be included in the model were chosen based on previous studies and known pathways of risperidone metabolism and response. For example, concomitant medications affecting CYP2D6 function (since risperidone is a CYP2D6 substrate); age and starting dose (as these are associated with AEs); race (as this may confound genotype); and sex (as the cohort was predominantly male) were included in the multivariable models.^3,7,19

Several additional exploratory analyses were performed post hoc. To look at the specific types of the most common AEs for each candidate gene, each AE type experienced by 4 or more individuals was tested for association for each gene using a Fisher’s exact test. A second exploratory analysis examined the association of dose on AE outcome including only those individuals started on a low dose of risperidone (0.5 mg/day or below); this analysis included a Fisher’s exact and a multivariable logistic regression of AEs by CYP2D6 metabolizer status adjusting for age, gender, and race. A third exploratory analysis was performed to examine the role of diagnosis as a covariate and in this model the most common diagnoses were included as a covariate in the final multivariable logistic regression model.

RESULTS

Study Cohort

Initially, 520 risperidone exposed children were identified in the biobank, and 270 (52%) met inclusion criteria. During genotyping 13 individuals were excluded due to ambiguous CYP2D6 metabolizer status. In the final cohort of 257 individuals, 188 (73%) were male, 217 (84%) were white, 246 (96%) were non-Hispanic, and median age on risperidone initiation was 8.3 (IQR 6.3–10.5) years (Table 1). The two most common psychiatric diagnoses in the cohort were ADHD (128, 50%) and autism (78, 30%).

Table 1:

Demographics, metabolizer status, and risperidone exposures in study cohort.

	N=257
	n (%)
Age at commencement of risperidone (years), median (IQR)	8.3 (6.3–10.5)
Sex
Male	188 (73.2)
Female	69 (26.8)
Race
White	217 (84.4)
African-American	29 (11.3)
Asian/Pacific Islander	5 (1.9)
Unknown	5 (1.9)
Native American	1 (0.4)
Ethnicity
Hispanic	6 (2.3)
Non-Hispanic	246 (95.7)
Unknown	5 (1.9)
Risperidone baseline dose in mg/day, median (IQR)	0.5 (0.5–1)
Risperidone maximum dose in mg/day, median (IQR)	1 (0.5,1.5)
Indication
Aggression	125 (48.6)
Behavioral problems	80 (31.1)
Agitation	42 (16.3)
Irritability	63 (24.5)
Self-injurious behaviors	45 (17.5)
Diagnosis
ADHD	128 (49.8)
Autism	78 (30.3)
Mood disorder	34 (11.6)
Tics	20 (7.8)
Depression	20 (7.8)
Time of risperidone treatment	200 (28, 657)
Individuals with an adverse event	76 (29.6)
Time to adverse event from start date (days) (IQR)	229 (73–479)
Time to adverse event from dose increase (days) (IQR)	83 (8–270)

IQR – Interquartile Range

Genotype

Table 2 includes the allele frequencies for the variants tested in DRD2, DRD3, HTR2A, and HTR2C; the metabolizer classification for CYP2D6, CYP3A4, and CYP3A5; and wild type status incorporating multiple or uncommon variants for ABCG2, ABCB1 and HTR2C. Sequencing of DRD2 and HTR2C failed to generate variant calls for a small number of individuals; statistical analyses for these genes were performed using only individuals with known genotypes.

Table 2:

Genotype frequencies by adverse event presence or absence.

	Individuals with No Adverse Event (n=181)	Individuals with Adverse Event (n=76)	P Value Univariate Analysis*		P Value Multivariate Analysis†
	Frequency (%) or median (IQR)
CYP2D6			0.04		0.03
PM/IM	18 (9.9)	15 (19.7)
NM/RM/UM	163 (90.1)	61 (80.3)
HTR2A (rs6311)			0.1		0.02
Wild Type	60 (33.2)	35 (46.1)
Heterozygote	95 (52.5)	35 (46.1)
Homozygote	26 (14.4)	6 (7.9)
ABCB1 (rs1045642, rs1128503, rs2032582)			0.6		0.2
Wild Type	154 (85.1)	62 (81.8)
Non-Wild Type (at least one variant)	27 (14.9)	14 (18.4)
ABCG2 (rs2231142)			0.7		0.8
Wild Type	144 (80.0)	59 (77.6)
Non-Wild Type	37 (20.6)	17 (22.4)
CYP3A4			0.6		1.0
PM/IM	11 (6.1)	5 (6.6)
NM/RM/UM	170 (93.9)	71 (93.4)
CYP3A5			0.6		0.1
PM/IM	170 (93.9)	70 (92.1)
NM/RM/UM	11 (6.1)	6 (7.9)
DRD2 (rs6277)			0.8		0.9
Wild Type	55 (30.4)	20 (26.3)
Heterozygote	81 (44.8)	37 (48.7)
Homozygote	41 (22.7)	18 (23.7)
Unknown	4 (2.2)	1 (1.3)
DRD2 (rs1799978)			0.8		0.2
Wild Type	137 (75.6)	65 (85.5)
Heterozygote	29 (16.1)	6 (7.9)
Homozygote	0	0
Unknown	15 (8.3)	5 (6.5)
DRD3 (rs6280)			0.5		0.5
Wild Type	26 (14.4)	14 (18.4)
Heterozygote	92 (50.8)	33 (43.4)
Homozygote	63 (34.8)	29 (38.2)
HTR2C (rs1414334)			0.7		1.0
Wild Type	29 (16)	12 (15.8)
Heterozygote	21 (11.6)	6 (7.8)
Homozygote	131 (72.4)	58 (76.3)
HTR2C (rs3818929, rs6318)			0.4		0.4
Wild Type	94 (51.8)	45 (59.2)
Non-Wild Type (at least one variant)	86 (47.5)	31 (40.8)
Unknown	1 (0.6)	0

^*P values among individuals with known genotypes from Fisher’s exact test

^†P values among individuals with known genotypes from logistic regression analysis adjusting for risperidone starting dose, age, sex, and race

Adverse Events

A total of 104 adverse events (AEs) were experienced by 76 (30%) individuals. In sum, 20 types of AEs were identified, and the most common included weight gain, sedation, and extrapyramidal symptoms (Table, Supplemental Digital Content 2). We have previously reported individual-level data regarding the specific AEs and clinician response.²⁰ In univariate analyses, CYP2D6 was the only gene associated with presence of AEs; PMs/IMs experienced more AEs than NMs/RMs/UMs (15/33, 46% vs. 61/224, 27%, P=0.04). Multivariate analyses were also performed for each gene, adjusting for risperidone starting dose, age, sex, and race. In the multivariable analyses, two genes were associated with AEs. CYP2D6 PMs/IMs had more AEs than NMs/RMs/UMs (OR 2.4, 95% CI 1.1–5.1), and individuals homozygous for HTR2A (rs6311) experienced fewer AEs than heterozygotes or wild types (OR 0.6, 95% CI 0.4–0.9).

Genes with significant associations (CYP2D6 and HTR2A) were included in a final multivariate model adjusting for risperidone starting dose, age, sex, and race. Results are shown in Figure 1 and remained consistent with the single-gene analyses, as CYP2D6 PMs/IMs had a higher odds ratio of AEs (OR 2.6, 95% CI 1.1–5.5) and individuals with the variant HTR2A rs6311 experienced fewer AEs (OR 0.6, 95% CI 0.4–0.9). Race was also significant in this model, with individuals of non-white background experiencing more AEs (OR 1.7, 95% CI 1.0–2.7).

Figure 1:

Final logistic regression model. For each variable, the following was associated with a numerically higher odds ratio for adverse events: poor/intermediate CYP2D6 metabolizer status, wild type HTR2A rs6311, increasing age, non-white race, lower starting dose, and male sex. Only CYP2D6, HTR2A, and race were statistically significant.

PM-Poor metabolizers

IM-Intermediate metabolizers

NM-Normal metabolizers

RM-Rapid metabolizers

UM-Ultrarapid metabolizers

With respect to our exploratory analyses, some genes were identified as having an association to a specific AE type. Notably DRD3 (rs6280) was associated with hypersalivation and hyperprolactinemia while DRD2 (rs1799978) was associated with hyperprolactinemia (Table, Supplemental Digital Content 3). When restricting to a low dose of initial risperidone dose (0.5 mg per day or less), the association of increased AEs in the CYP2D6 PMs/IMs vs. NMs/RMs/UMs group remained significant in univariate analysis (9/15, 60%, vs. 33/123, 26.8%, p=0.02) and multivariable analysis (OR 4.5, 95% CI 1.4–14.2). For the third exploratory analysis, both CYP2D6 and HTR2A remained significant when the multivariable model included adjustment for diagnosis, dose, age, gender, and race. In the model including the diagnosis of ADHD, individuals with HTR2A rs6311 experienced fewer AEs (OR 0.6, 95% CI 0.4–0.9) and CYP2D6 PMs/IMs had a higher odds ratio of AEs (OR 2.5, 95% CI 1.2–5.6); in this model race was also significant, with individuals of non-white background experiencing more AEs (OR 1.7, 95% CI 1.0–2.7) while diagnosis of ADHD was not significant (OR 1.1, 95% CI 0.7–2.0). When this model was run for the diagnosis of autism both CYP2D6 and HTR2A remained significant; individuals with HTR2A rs6311 experienced fewer AEs (OR 0.6, 95% CI 0.4–0.9) and CYP2D6 PMs/IMs had a higher odds ratio of AEs (OR 2.6, 95% CI 1.1–2.6); in this model race was also significant, with individuals of non-white background experiencing more AEs (OR 1.6, 95% CI 1.0–2.7) while diagnosis of autism was not significant (OR 1.0, 95% CI 0.6–2.0).

DISCUSSION

Clinical pharmacogenomic testing is available from a variety of commercial and academic laboratories, and some tests include recommendations for genotype guided dosage. However, there are no formal guidelines for risperidone prescription guidance based on genotype. The results of this study demonstrate that children with poor or intermediate CYP2D6 metabolism experienced more adverse events (AEs) than to those with normal, rapid, or ultrarapid metabolism. These results were consistent in univariate and multivariate analyses and are concordant with previous reports. Results also showed association of fewer AEs in homozygotes for rs6311 (HTR2A) than in heterozygotes or wild type individuals. The other genes tested in this study (ABCB1, ABCG2, CYP3A4, CYP3A5, DRD2, DRD3, and HTR2C) did not show an association of risperidone AEs in our pediatric cohort. This study represents the largest sample examining the association of each variant to pediatric risperidone AEs and represents the only study that combines this number of genetic variants with clinical data to predict risk of pediatric risperidone AEs.

The association of increased AEs in risperidone exposed pediatric patients with reduced CYP2D6 activity is consistent with multiple prior pediatric studies (Table 3). Four prior studies have shown increased plasma risperidone or AEs, such as weight gain and abnormal movements, in children with reduced CYP2D6 metabolism during risperidone exposure.^21–24 Another small study showed reduced weight gain in ultrarapid CYP2D6 metabolizers during risperidone treatment.²⁵ Taken together, these studies and our data suggest increased AEs in pediatric CYP2D6 PMs/IMs and support CYP2D6-guided prescribing of risperidone.

Table 3:

Summary of 11 prior pediatric risperidone pharmacogenomic studies totaling 848 children.

Study	Genes and Variants	Outcomes	Number of Participants	Age of Participants	Results
Youngster, et al.21	CYP2D6 *2, *3, *4, *5, *6, *8, *9, *10, *11, *14, *15, *17, *18, *19, *20, *25, *26, *29, *30, *31, *35, *36, *37, *40, *41, *43, *52, and gene duplication	Weight gain and neurological events	40	3–18 years	AEs in 2/2 PMs, 9/35 IMs/NMs, and 0 UMs, but non significant association of PMs vs IMs/NMs (P=0.08) and UMs vs IMs/NMs (p=1.0); however, PMs had higher risperidone levels (p=0.03) and higher ratios of risperidone to metabolite (p=0.02) compared to IMs/NMs
Nussbaum, et al.22	CYP2D6 *4	Weight gain	81	9–20 years	IMs with more weight gain than NMs/RMs/UMs (p<0.001)*
Dos Santos-Júnior, et al.23	CYP2D6 * 10	Weight gain, hypertension, and metabolic abnormalities	120	8–20 years	Increased hypertension (p=0.039) and abdominal circumference (p=0.016) in PMs vs IMs/NMs/RMs/UMs
Grădinaru R, et al.24	CYP2D6 *3, *4, *5, *41	Prolactin and clinical adverse events	81	9–20 years	IMs with increased prolactin during treatment vs NMs (p<0.001)*
Correia, et al.25	CYP2D6 *3, *4, *5, *6, gene duplication	BMI, waist circumference, neurological events	45	3–21 years	UMs associated with BMI or waist circumference increase vs NMs (p=0.002)
Correia, et al.24	HTR2A rs6311	BMI, waist circumference, neurological events	45	3–21 years	Wild type with increased prolactin (p=0.006) vs non-wild type
Rafaniello, et al.7	CYP3A4 *22 CYP3A5*3	Liver enzymes and adverse events	64	Below 18 years	No association
Correia, et al.25	DRD3 rs6280	BMI, waist circumference, neurological events	45	3–21 years	No association
Rafaniello, et al.7	ABCB1 rs1045642 and rs2032582	Liver Enzymes, creatinine, and adverse events	64	Below 18 years	No association
Correia, et al.25	ABCB1 rs1128503 and rs1045642	BMI, waist circumference, neurological events	45	3–21 years	No association
Sukasem, et al.26	ABCB1 rs2032582 and rs1045642	Insulin resistance	89	3–20 years	No association
Dos Santos-Júnior, et al.23	HTR2C rs1414334, rs6318 and rs3813929	Weight gain and metabolic abnormalities	120	8–20 years	rs6318 heterozygotes with increased abdominal circumference in females compared to wild type/heterozygotes (p=0.035); rs3813929 heterozygotes with increased leptin compared to hemi/homozygotes (p=0.044)
Correia, et al.25	HTR2C rs6318, rs3813928, and rs3813929	BMI, waist circumference, neurological events	45	3–21 years	rs6318 wild type women/hemizygous men with lower prolactin increase than homo/ hemizygotes (p=0.006); rs6318 homozygous women and male hemizygotes with increased waist circumference (p<0.01) and BMI (p=0.037) compared to wild type
Hoekstra, et. Al19	HTR2C rs1414334 and rs3813929	BMI	32	5–16 years	rs3813929 homo/heterozygotes with reduced weight gain compared to wild type (p<0.001)
Del Castillo, et. al27	HTR2C rs3813929	BMI	45	7–17 years	No association
Almandil, et. al28	HTR2C rs1414334	BMI	144	2–17 years	No association
Rafaniello, et al.7	ABCG2 rs2231142	Liver Enzymes, creatinine, and adverse events	64	Below 18 years	Variant homozygotes with higher probability of nutrition and metabolism disorders (p=0.008) vs heterozygotes/ wild type*
Dos Santos-Júnior, et al.23	DRD2 rs6277 and rs1799978	Weight gain and metabolic abnormalities	120	8–20 years	rs1799978 heterozygotes with increased insulin resistance compared to wild type (p=0.010); Homozygotes of rs6277 with increased insulin resistance compared to wild type (p=0.053)
Calarge, et al.29	DRD2 rs6277, rs1800497, rs1799732, rs 1799978	Prolactin	107	7–17 years	rs1799978 hetero/homozygotes with increased prolactin elevation compared to wildtype in multivariable model (p=0.002); rs1800497 allele carriers with increased prolactin elevated compared to wild type (p=0.04)

^*Included other antipsychotics

PM-Poor metabolizers

IM-Intermediate metabolizers

NM-Normal metabolizers

RM-Rapid metabolizers

UM-Ultrarapid metabolizers

Our finding of reduced AEs in homozygotes for rs6311 in the HTR2A gene is consistent with one prior pediatric study, which specifically identified increased prolactin as an AE in wild type compared to those with the rs6311 variant in HTR2A (Table 3).²⁵ Hyperprolactinemia in children can result in a variety of concerns including amenorrhea, gynecomastia, and hypogonadism.²⁴ As our study was retrospective and included data from a variety of providers, not all individuals had prolactin measurements. However, in our cohort, 4 individuals did experience hyperprolactinemia or gynecomastia; 3 of these individuals had the variant while one was wild type. It is possible that there is an association with this variant and hyperprolactinemia, but due to the small numbers of hyperprolactinemia detected for this study we were not able to replicate this association as demonstrated in our exploratory analyses of specific AEs listed in Table, Supplemental Digital Content 3. Considering the limited number of studies on this variant in pediatrics, changes in clinical practice (e.g. personalized risperidone dosing) based on genetic testing for rs6331 in the HTR2A are not warranted at this time but should be further examined in larger studies.

For CYP3A4, CYP3A5, DRD3, and ABCB1, our results are consistent with all prior literature reporting no association of genetic variants to risperidone AEs in children (Table 3). To our knowledge, for these variants, there are no prior studies that have found an association with risperidone AEs. There are reports of negative findings when the association of CYP3A4, CYP3A5, and DRD3 (rs6280) with AEs, consistent with our data.^13,25 It is important to note, that in exploratory analyses in our study, DRD3 (rs6280) was associated with specific subtypes of AEs with the variant allele with an apparent protective effect with respect to hyperprolactinemia/gynecomastia/precocious puberty and hypersalivation. Three prior studies have also failed to show an association of variants of ABCB1 (including rs1045642, rs1128503, and rs2032582) with risperidone-induced AEs.^7,25,26 Taken together, CYP3A4, CYP3A5, DRD3, and ABCB1 genetic test results should not be used to guide risperidone prescribing in the pediatric population at this time although larger studies examining specific AE subtypes are warranted, particularly examining DRD3. These negative associations should be kept in mind when interpreting pharmacogenetic reports, which may include variants in these genes despite the lack of evidence.

HTR2C, ABCG2, and DRD2 have mixed pediatric evidence with some prior studies reporting an association and others reporting no significant association (Table 3). There are three prior pediatric studies that have shown an association between variants in HTR2C and weight gain in children, but our study did not show an association of all AEs or weight gain only and these same variants.^19,23,25 There are also two other pediatric studies that looked at the same variants of HRT2C that showed no association with AEs.^27,28 There is only one small prior study examining ABCG2 (rs2231142), which showed an association with metabolic and nutrition AEs during risperidone treatment while our results showed no association to all AEs or metabolic only AEs (Table, Supplemental Digital Content 3).⁷ One prior study in children and adolescents demonstrated elevated insulin resistance with the DRD2 rs6277 variant which did not show an association with AEs in our study.²³ However, another study of DRD2 rs6277 showed an association of this variant with elevated prolactin.²⁹ In exploratory analyses of our study, individuals who were non wildtype for DRD2 1799978 did have increased odds hyperprolactinemia/gynecomastia/precocious puberty. Thus, considering the smaller sample sizes of these prior studies, and the overall inconsistent pediatric findings, there is insufficient pediatric data to support using HTR2C, ABCG2, and DRD2 for risperidone prescription guidance for children at this time, although specific risperidone AE subtypes such as hyperprolactinemia requires further investigation. Again, these negative findings may be helpful when interpreting pharmacogenetic reports.

The exploratory analysis examining AEs in only those started on a low dose should be interpreted with caution in the setting of a retrospective cohort study. Although our results showed a continued increase of AEs in CYP2D6 PMs/IMs despite starting on a lower dose in univariate and multivariable analysis, it is possible that there is selection bias. It is likely those individuals that clinicians felt were especially prone to AEs (such as those with an autism spectrum disorder) were started on a lower dose. Thus, while these findings suggest that starting at a lower dose might not mitigate the effect of increased AE risk due to genotype, this association should be further investigated in prospective studies.

Other considerations should be taken into account during risperidone prescribing, which include concomitant medications, age, clinical indication, medical comorbidities, and race. Race was also associated with AEs in the final multivariable analysis model in our study. Prior literature has reported differences in CYP2D6 polymorphisms across ancestries.³⁰ Other considerations that could affect the association of race that were not available in our data include socioeconomic factors.

There are several limitations to this study. Data were from a single tertiary care center and might not be generalizable to other pediatric populations. AE identification was performed retrospectively through chart review for patients seen by a variety of providers, and thus identification of AEs might be limited by inaccurate or insufficient documentation. It was also not possible to determine medication adherence from the EHR data. As this was a retrospective study, no causality assessment between AEs and risperidone was performed, although clinically each clinician, caregiver or patient attributed the AEs to risperidone. This study outcome was limited to AEs. It is possible that other outcomes, such as response, are also affected by the genes tested and could provide valuable information for risperidone prescribing. Although our sample size was relatively large, it was not possible to stratify by metabolizer subgroups for CYP2D6, CYP3A4, and CYP3A5. Thus, it was not possible to assess each phenotype individually or identify trends among classes.

The results of this study indicate that children who are CYP2D6 poor or intermediate metabolizers and children without the HTR2A-rs6311 variant may have increased risk for AEs during risperidone treatment. Overall, CYP2D6 has the most robust evidence supporting its use for pharmacogenomic risperidone guidance. If a child is a known CYP2D6 poor or intermediate metabolizer, clinicians could consider using alternative medications, reducing the risperidone starting dose, performing slower risperidone titration or increasing AE monitoring. Many of the other genetic markers in this study, as well as others reported on commercial laboratory pharmacogenomic testing, have mixed or no pediatric data, and thus are not clinically actionable at this time.