Cytochrome P450 2D6 *17 and *29 Allele Activity for Risperidone Metabolism: Advancing Precision Medicine Health Equity

Abstract

CYP2D6 alleles with low frequency in Eurocentrically biased study populations are often excluded from pharmacogenetic investigation and consequently may have misassigned activity values. This health inequity may be contributing to imprecise dose predictions for CYP2D6‐metabolizing drugs. The objective of this study was to determine how sub‐Saharan African‐specific CYP2D6*17 and *29 alleles affect risperidone metabolism. To do this, we generated the largest real‐world cohort of risperidone users in an African study population for pharmacogenetic studies. Risperidone users ≤ 18 years old were recruited from the Federal Neuro‐Psychiatric Hospital. Health records were obtained by parent report and paper charts. CYP2D6 genotyping was performed for > 20 variants. Plasma concentrations of risperidone and 9‐hydroxyrisperidone were determined by liquid chromatography mass spectrometry. CYP2D6 activity was calculated based on the metabolic ratio 9‐hydroxyrisperidone:risperidone. Multivariable linear regression modeling was performed to determine the association between our alleles of interest and log‐transformed ratio‐defined CYP2D6 activity relative to star alleles with established activity values. Across 208 enrolled participants, CYP2D6 activity value for *17 was found to be twice that of normal function alleles, while *29 was comparable to no function alleles. These results contrast previous values assigned to *17 and *29 from guidelines, which are not based on evidence with risperidone, suggesting the possibility of substrate specificity for these alleles. Ultimately, our findings have the potential to improve risperidone prescribing, especially for patient groups with substantial sub‐Saharan African ancestry. Importantly, this work underscores the critical need to better understand the effects of ancestry‐specific alleles for achieving equitable pharmacotherapeutic health outcomes.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Data on the metabolic activity of the CYP2D6*17 and *29 alleles for risperidone and other CYP2D6 substrates is lacking. This paucity in evidence may be contributing to misassigned activity values and therefore imprecise dose predictions for risperidone, especially for patient groups historically excluded from pharmacogenetic research participation.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This study addresses the metabolic activity of CYP2D6*17 and *29 alleles for risperidone and if this activity is associated with risperidone adverse drug reactions

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study suggests that CYP2D6*17 demonstrated increased metabolic activity for risperidone, while the *29 allele had very little activity, both in contrast to the current activity value of 0.5, indicating decreased function, determined by the Clinical Pharmacogenetics Implementation Consortium.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

This study suggests that clinical pharmacogenetic dosing guidelines for CYP2D6‐metabolized drugs should account for the substrate when assigning *17 allele activity values. Implementing this paradigm shift in CYP2D6 clinical pharmacogenetic implementation could not only improve pharmacotherapeutic outcomes but may also promote health equity.

Cytochrome P450 2D6 (CYP2D6) is a hepatic enzyme that contributes to the metabolism of approximately 20–25% of commonly prescribed medications, including antidepressants, antipsychotics, opioids, and beta‐blockers. CYP2D6 is highly polymorphic, and enzyme variation significantly influences differing drug concentrations. Guidelines from organizations such as the Clinical Pharmacogenetics Implementation Consortium (CPIC) and the Dutch Pharmacogenetics Working Group (DPWG) recommend using CYP2D6 test results if available to optimize drug prescribing for substrates of this enzyme. ¹ Moreover, randomized controlled trial data show that the implementation of CYP2D6 testing as part of a preemptive pharmacogenetic panel reduces the risk of adverse drug reactions. ² Consequently, CYP2D6 pharmacogenetic testing is increasingly performed in clinical settings to predict enzyme activity and tailor drug therapy for improved pharmacotherapeutic outcomes. ³

The prediction of CYP2D6 enzyme activity from pharmacogenetic testing is based on a scoring approach where preassigned activity values of both inherited star alleles are summed for an individual to give an activity score. ⁴ The activity values for individual alleles are routinely assessed by CPIC and can be changed when the evidence supports a change, as was done for the CYP2D6*9 and *41 alleles recently. ⁵ , ⁶ Understudied alleles are the most likely to have misassigned activity values ¹ . This scientific knowledge gap presents as a health equity issue because minoritized groups are less likely to be represented in genomic studies. ⁷ The most commonly cited reasons underlying this lack of representation are directly or indirectly rooted in racism at different levels (e.g., structural, institutional, interpersonal) including mistrust due to historically racist practices (e.g., the Tuskegee Syphilis Study), competing demands (e.g., employment, family care) that make participation inconvenient, and limited health/research literacy. ⁷ , ⁸ , ⁹ , ¹⁰ Addressing this knowledge gap will lead to more precise pharmacogenetic‐guided prescribing and in turn improved health outcomes for a broader patient population than currently observed. ¹

Risperidone is an atypical antipsychotic drug used in the treatment of schizophrenia, bipolar disorder, and autism‐related irritability in youth. CYP2D6 metabolizes risperidone (which is active) into an active metabolite, 9‐OH‐risperidone, the majority of which is excreted unchanged. ¹¹ Therapeutic drug monitoring guidelines indicate that the total active moiety (risperidone +9‐OH‐risperidone) is what is important for efficacy. ¹² However, there are differences in side effect profiles of risperidone and paliperidone (9‐OH‐risperidone) ¹³ , ¹⁴ and brain penetration is greater for risperidone than 9‐OH‐risperidone. ¹⁵ The DPWG recommends reduced risperidone doses in CYP2D6 poor metabolizers, ¹⁶ and a CPIC guideline is in progress (https://cpicpgx.org/genes‐drugs/).

Most studies investigating the effect of CYP2D6 on risperidone metabolism to date have been conducted primarily in Eurocentrically biased study populations, with no studies involving pediatric patients in Africa. The activity values assigned to the sub‐Saharan African‐specific CYP2D6*17 and *29 alleles are based on several substrates (debrisoquine, bufuralol, codeine, dextromethorphan, metoprolol, tamoxifen, primaquine) but do not include risperidone. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ One of these studies indicates there may be substrate‐specific effects, with decreased activity of the *17 and *29 alleles for debrisoquine, metoprolol, and dextromethorphan but not codeine. ²¹ CPIC’s assessment of the activity values of these alleles is currently rated as moderate, based on 4–6 studies in vitro and in vivo whereas the *10 allele is rated as strong based on 18 studies, and the *41 allele is rated as strong based on 5 studies. The primary objective of this study was to determine the effects of alleles *17 and *29 on CYP2D6‐mediated risperidone metabolism in a Nigerian pediatric population. We additionally explored the relationship between these predictors and risperidone‐associated adverse drug reactions.

MATERIALS AND METHODS

Participant recruitment and enrollment

Participants were recruited from the Child and Adolescent Mental Health Service Center of the Federal Neuro‐Psychiatric Hospital during their regularly scheduled outpatient appointments. For study inclusion, participants had to be: (i) ≤ 18 years of age at the time of initial dose and (2) adherent to oral risperidone for > 2 weeks per routine care as reported by the patient or guardian. Patients with a history of traumatic brain injury, substance abuse, organ transplant, or congenital brain abnormality were excluded. Participants who met eligibility criteria and provided written informed consent (through their guardians) were enrolled in the study.

Sample and data collection

At the time of study enrollment, 8 mL of whole blood was collected from each participant into two ethylenediaminetetraacetic acid‐coated tubes. A portion of the whole blood (6 mL) was spun at 2,000 revolutions per minute for 10 minutes (room temperature) to extract plasma. Whole blood (2 aliquots) and plasma samples (4 aliquots) were frozen within 15 minutes of spinning at −80°C for future analyses.

Demographic and clinical data were obtained manually from patient paper medical records or directly from a guardian. Data included demographics (age, sex, and ethnicity), risperidone prescribing history (indication, initiation date, dosage, time since last dose, and past dose changes or discontinuation), risperidone‐associated adverse drug reactions (weight changes, drowsiness, hyperprolactinemia, drooling, extrapyramidal effects, and appetite changes), and other relevant information (comorbidities and concurrent medication use including CYP2D6 inhibitors). For ethnicity, patients or guardians could select from any of the following options: Yoruba, Igbo, Hausa, Fulani, and Other (as a write‐in entry).

Genotyping and genotype‐defined CYP2D6 enzyme activity scores

Deoxyribonucleic acid (DNA) was isolated from whole blood for genotyping using the Perkin Elmer Chemagic 360 and following the manufacturer’s instructions. A custom Agena MassARRAY® assay (Agena Bioscience, San Diego, CA) was used for genotyping. The test detects 22 different alleles in CYP2D6 that are associated with no function (*3, *4, *5, *6, *7, *8, *11, *12, *15, *18, *19, *20, *36, *69, *114), decreased function (*9, *10, *14, *17, *29, *41), or normal function (*2). A *1 allele was inferred if none of the 22 above variant alleles were present. Variants tested are provided in Table S1 . Though it is not possible to phase the alleles with this system, the most likely allele (based on population data) is reported, as recommended by the American College of Medical Genetics and Genomics. ²⁵ Selected samples (n = 12) were sequenced using a variety of orthogonal methods, and all alleles called by the MassARRAY® assay were confirmed, with no new missense variants discovered. The VeriDose® CYP2D6 CNV panel of the MassARRAY® system was used for CYP2D6 copy number detection. Metabolizer phenotypes (poor metabolizers [PM], intermediate metabolizers [IM], normal metabolizers [NM], and ultrarapid metabolizers [UM]) were predicted from activity scores assigned to allelic combinations in accordance with standard recommendations. ⁴ This includes activity values of 0.5 for our alleles of interest (*17 and *29). Phenoconversion of participants taking a concomitant CYP2D6 inhibitor (n = 4 total, 3 fluoxetine and 1 quinidine) was not considered as we did not collect the dose or timing of administration of the inhibitor. We performed a sensitivity analysis with these patients classified as poor metabolizers instead of normal metabolizers, which did not change our results or interpretation (Figure S1 ).

Quantification of drug concentrations and ratio‐defined CYP2D6 enzyme activity scores

Plasma concentrations of risperidone and 9‐hydroxyrisperidone (an active risperidone metabolite) were analyzed by tandem liquid chromatography mass spectrometry. Specifically, extraction was performed by aliquoting 25 μL of control plasma or sample to polypropylene microcentrifuge tubes. Standards received 25 μL of spiking standard, with blanks and samples receiving 25 μL of acetonitrile. Samples and standards were then spiked with 25 μL of internal standard, with blanks receiving 25 μL of acetonitrile. Hundred microliters of methanol was added to all samples prior to being vortexed to mix and centrifuged. Thereafter, 100 μL of the resulting supernatant was transferred to a 96 polypropylene collection tube along with 100 μL of methanol. Extracts were analyzed on a Waters Xevo TQ‐XS Triple Quadrupole Mass Spectrometer with a Waters Acquity FTN‐I (ultra‐performance liquid chromatography I Class Plus) in positive mode using a Waters Acquity ultra‐performance liquid chromatography BEH C18, 1.7 μm, 2.1 × 50 mm column. The samples were injected with an isocratic gradient of 85:15 MPH A (100:0.1 H2O:HCOOH):MPH B (100:0.1 ACN:HCOOH). Transitions of 411 > 191, 415 > 194, 427 > 206, and 431 > 210 were monitored for risperidone, risperidone‐d4, 9‐hydroxyrisperidone, and 9‐hydroxyrisperidone‐d4, respectively. The calibration curve range of detection for the assay was 0.1–100 ng/mL for both analytes.

We used the dose‐adjusted metabolite:parent (9‐hydroxyrisperidone:risperidone) ratio as a biomarker of CYP2D6 enzyme activity. From this biomarker, we calculated the ratio‐defined CYP2D6 activity as previously described. ²⁶ Briefly, we set the median normalized metabolic ratio of patients carrying the *1/*1 diplotype to 1.0 and those carrying two no function alleles (i.e., poor metabolizers) to 0.0. Though it has normal function, we did not include *2 carriers in the normal function group because of discrepancies in its function across substrates. ²⁷ , ²⁸ The relative CYP2D6 activity score for each patient was calculated according to the equation below:

$$\left(\mathrm{MR}-x\right)/\left(y-x\right)$$

where MR is the individual patient’s time‐since‐last‐dose‐adjusted metabolite:parent ratio (9‐hydroxyrisperidone:risperidone), x is the median metabolic ratio of the poor metabolizers, and y is the median metabolic ratio of the patients with a *1/*1 diplotype. Note that the term “activity value” is used to denote the CPIC‐defined activity value, while our experimentally determined activity is termed “ratio‐defined CYP2D6 activity.”

Statistical analyses

We conducted a series of regression analyses.

To determine the effect of CYP2D6 variation on pharmacokinetic parameters of risperidone metabolism, we used linear regression. In particular, concentrations (dose‐adjusted risperidone, dose‐adjusted 9‐hydroxyrisperidone, and dose‐adjusted risperidone +9‐hydroxyrisperidone [total active moiety]) as well as ratios (metabolite:parent, parent:metabolite) and most recent risperidone dose were log‐transformed prior to analyses and used as dependent variables. Among our dependent variables, metabolite:parent ratio (our biomarker of CYP2D6 activity) was considered to be the primary endpoint. The prespecified independent variable was CYP2D6 genotype‐defined activity score or phenotype, whichever fit the given model better. Covariates included age, sex, ethnicity, most recent risperidone dose (only for ratios since other dependent variables were already dose‐adjusted), and hours after the last risperidone dose. Moreover, a subsequent analysis – with genotype‐defined activity score substituted out for CYP2D6 allele activity value as the new independent variable – was conducted to determine the effect of individual alleles. Based on the effect of individual alleles in our subsequent analysis, we reassigned activity values for CYP2D6 *17 and *29. These new activity values were used to generate revised CYP2D6 genotype‐defined activity scores followed by a repeat regression of genotype‐defined activity scores on the primary endpoint with the same covariates as above.

To determine the effect of CYP2D6 on risperidone‐associated adverse drug reactions, we used logistic regression. For this set of analyses, the dependent variable could include dose reduction (all‐cause), weight gain, appetite change, hyperprolactinemia, drooling, extrapyramidal symptoms, drowsiness, or any adverse drug reaction. Prespecified independent variables included age, sex, weight, ethnicity, dose, CYP2D6 phenotype, CYP2D6 genotype‐defined activity scores, revised CYP2D6 genotype‐defined activity scores, ratio‐defined CYP2D6 activity, risperidone concentration, 9‐hydroxyrisperidone concentration, and total active moiety.

Multivariable models for all regression analyses were constructed with only the variables/covariates associated with the dependent variable (P < 0.05) following stepwise backward elimination to avoid collinearity. Best‐fitting multivariable models were determined using the Akaike information criterion (AIC) for logistic regression and R ² for linear regression. R version 4.4.0 was used for analysis.

Ethics approval

Ethical approval to conduct the study was obtained from the Human Research and Ethics Committee of the Federal Neuro‐Psychiatric Hospital, Lagos, Nigeria, and the University of California, San Francisco Institutional Review Board.

RESULTS

Demographic and clinical characteristics of the study population

A total of 208 enrolled participants had CYP2D6 genotyping data available for analysis (Figure 1 ). Participants ranged in years of age from 2 to 18 at the time of sample collection (with a mean age of 11 years). The proportion of participants assigned female at birth was 35%. The most common diagnosis among participants was attention deficit hyperactivity disorder (ADHD). Multiple ethnic groups were represented in the study population including 118 of Yoruba ethnicity (57%) and 54 of Igbo (26%). See Table 1 and Table S2 for more demographic and clinical characteristics. Age and dose were significantly correlated, but age did not differ across CYP2D6 metabolizer groups (Figure S2 ).

Figure 1.

Cohort diagram showing how many patients were included in each analysis. ADRs, adverse drug reactions. PK, pharmacokinetics; Risp, risperidone; 9‐OH‐Risp, 9‐hydroxyrisperidone.

Table 1.

Clinical characteristics of participants

	Total (N = 211)
Sex
Female	74 (35.1%)
Male	137 (64.9%)
Age at starting risperidone (years)
Mean (standard deviation)	8.5 (4.5)
Range	1.0–17.0
Age at time of sample (years)
Mean (standard deviation)	11.3 (4.8)
Range	2.0–18.0
Weight (kg)
Mean (standard deviation)	40.5 (20.1)
Range	11.0–105.0
Ethnicity
Edo/Esan	9 (4.3%)
Igbo	54 (25.6%)
Other	27 (12.8%)
Yoruba	121 (57.3%)
Indications for risperidone use
Depressive Disorder	2 (0.9%)
ADHD (secondary symptoms)	81 (38.4%)
Autism	60 (28.4%)
Bipolar	6 (2.8%)
Schizophrenia	26 (12.3%)
CYP2D6 Inhibitor	4 (1.9%)
Risperidone dose (mg/day)
Mean (standard deviation)	2.3 (2.2)
Range	0.125–10.0
Days on risperidone
Median (range)	589 (9–4,838)

ADHD, attention‐deficit/hyperactivity disorder.

CYP2D6 genetic variation

Genotyping revealed a diverse range of CYP2D6 alleles within the study population. In particular, the *17 and *29 sub‐Saharan African‐specific alleles had frequencies of 19% and 11%, respectively (Table 2 , Table S3 ). Accordingly, CYP2D6 genotype‐defined activity scores and corresponding metabolizer phenotype varied substantially. Genotype‐defined activity scores ranged from 0 to 3.0 (median 1.5, interquartile range 1.0 to 2.0). There were a total of 119 participants (57%) classified as NM, 7 as PM (3%), 72 as IM (35%), and 10 as UM (5%).

Table 2.

Allele frequency and activity calculations.

Allelea	Count in cohortb	Count for activityc	% Unused for activity	Activity model estimated	Ratio to *1e	Current activity valuef
	(N, %)	(n)	([N−n]/N)	Estimate (SE)
*1	134 (31.8%)	83	38%	0.490 (0.504)	1	1
*1 × 2	3 (0.7%)	0	100%	NA	NA	2
*2	62 (14.7%)	42	32%	0.370 (0.557)	0.76	1
*2 × 2	7 (1.7%)	4	43%	1.171 (0.859)	2.39	2
*4	4 (0.9%)	3	25%	−0.938 (0.839)	−1.91	0
*4 × 2	11 (2.6%)	9	18%	−0.390 (0.649)	−0.8	0
*4 × 5	1 (0.2%)	1	0%	NA	NA	0
*5	36 (8.5%)	25	31%	−0.763 (0.557)	−1.56	0
*10	14 (3.3%)	8	43%	0.162 (0.642)	0.33	0.25
*17	80 (19%)	33	59%	1.227 (0.503)	2.5	0.5
*17 × 2	2 (0.5%)	1	50%	NA	NA	1
*29	48 (11.4%)	38	21%	0.075 (0.516)	0.15	0.5
*29 × 2	1 (0.2%)	1	0%	NA	NA	1
*40	7 (1.7%)	5	29%	−0.235 (0.748)	−0.48	0
*41	2 (0.5%)	1	50%	NA	NA	0.25
*41 × 2	1 (0.2%)	1	0%	NA	NA	0.5
*68 + *4	2 (0.5%)	2	0%	NA	NA	0
*68 + *4 × 2	1 (0.2%)	1	0%	NA	NA	0
Unknown	6 (1.4%)	NA	NA	NA	NA	NA

SE, standard error.

^a Alleles included in genotyping assay but not observed: *3, *6, *7, *8, *9, *11, *12, *13, *14, *15, *18, *19, *20, *36, *69, *114.

^b Indicates the number of alleles observed in the full study population (211 total participants for N = 422 total alleles).

^c Indicates the number of alleles among the full study population with a sample within 48 hours of the last risperidone dose (187 participants), detectable levels for both the parent and metabolite concentration (130 participants), and successful CYP2D6 genotyping (129 participants for n = 258 total alleles).

^d Activity was estimated for each patient, then a linear model was used with log(activity) as the dependent variable, and each allele with at least 3 observations was included as variables in the model, along with hours since the last dose (among n = 129). One patient included in the activity estimate did not have the risperidone dose recorded. NA is shown for alleles with fewer than three observations for activity.

^e Ratio to *1 is calculated based on the “Activity model estimate” values to provide a percentage of normal enzyme function for comparison with the Current Activity Value.

^f Based on consensus recommendations from the Clinical Pharmacogenetics Implementation Consortium and Dutch Pharmacogenetics Working Group. ⁴

Ratio‐defined CYP2D6 enzyme activity vs. genotype‐defined scores

There was also substantial variation across the study population for ratio‐defined CYP2D6 activity, as expected. Ratio‐defined activity ranged from 0.0 to 12.7 (median 0.7, interquartile range 0.3 to 1.6). Genotype‐defined CYP2D6 phenotypes generally demonstrated substantial intragroup variability in ratio‐defined CYP2D6 enzyme activity (IM range 0.06 to 5.5; NM range 0.02 to 12.6; UM range 1.2 to 12.7), though PMs all had very little activity (range − 0.02 to 0.05). Intergroup comparisons for ratio‐defined CYP2D6 enzyme activity scores showed an increasing gradient across the PM–IM–NM–UM spectrum (median scores of 0.0, 0.62, 0.74, 2.10, respectively) consistent with the expected order of metabolic capacity. Notably, however, median scores for the IM and NM phenotype groups were more similar than expected (Figure 2 a ).

Figure 2.

Ratio‐defined CYP2D6 activity compared to genotype‐defined CYP2D6 activity score. (a) Ratio‐defined CYP2D6 activity is plotted vs. current CYP2D6 activity score (P = 0.0002, adjusted R ² = 0.101). In a multivariable model including hours after the last dose and risperidone dose, current CYP2D6 activity score accounted for a change in adjusted R ² of 0.123. (b) Calculated CYP2D6 activity is plotted vs. proposed CYP2D6 activity score (changing activity value of *17 allele from 0.5 to 2 and *29 from 0.5 to 0, P < 0.00001, adjusted R ² = 0.232). In a multivariable model including hours after the last dose and risperidone dose, current CYP2D6 activity score accounted for a change in adjusted R ² of 0.299.

Ratio‐defined activity was associated with genotype‐defined activity scores in a univariate analysis (beta = 0.74, P = 0.0002, Table S4 ). Significant predictors of ratio‐defined activity were risperidone dose, hours after the last dose, genotype‐defined activity score, and CYP2D6 phenotype. The best‐fitting multivariable model for ratio‐defined activity included hours after the last dose, risperidone dose, and CYP2D6 activity score as covariates (R ² = 0.256, Table S4 ).

When a subsequent regression analysis was performed to determine the effect of individual CYP2D6 alleles, the effect for *17 (1.2) was found to be more than double that of the *1 and *2 normal function alleles (0.49 and 0.37, respectively, Table 2 ). Additionally, the mean effect for *29 (0.075) was found to be between that of the decreased function *10 (0.16) and no function alleles *4 and *5 (−0.94 and −0.76). The ratio of each allele to *1 is found in Table 2 , accompanied by the current CYP2D6 activity value determined by CPIC.

Ratio‐defined CYP2D6 enzyme activity scores vs. revised genotype‐defined scores

Based on the observed effects for our alleles of interest relative to those from well‐studied alleles, we recalculated activity scores for each patient by reassigning activity values for *17 (changed from 0.5 to 2.0) and *29 (from 0.5 to 0.0). The association between ratio‐defined and revised genotype‐defined scores (beta = 0.68, P < 0.0001) was stronger compared to the unrevised scores. Revising the genotype‐defined scores also improved the multivariable model for ratio‐defined CYP2D6 activity (R ² = 0.432, Table S4). Of note, the unrevised activity score explained 10% of the variability in the ratio‐defined activity, while the revised activity score explained 23%.

Revising the genotype‐defined CYP2D6 activity scores also improved the gradient of ratio‐defined scores across the PM–IM–NM–UM spectrum (Figure 2 b ). Thus, the aforementioned similarity in ratio‐defined CYP2D6 activity between IM and NM groups was found to be attributed at least partially to our alleles of interest (CYP2D6 *17 and *29).

Other markers of risperidone metabolism

Risperidone, 9‐hydroxyrisperidone, total active moiety, and risperidone dose were also associated with the CYP2D6 metabolizer phenotype (Figure 3 , Tables S5–S9 ). Additionally, hours after the last dose were associated with both lower risperidone concentrations (P < 0.0001) and lower parent:metabolite ratio (P = 0.0001) with increasing slopes (i.e., higher clearance) across the PM‐IM‐NM‐UM spectrum (Figure S3 ).

Figure 3.

Association of CYP2D6 phenotype with risperidone and 9‐hydroxyrisperidone concentrations. (a) Association of dose‐adjusted risperidone concentrations with CYP2D6 phenotype, P = 0.02; (b) Association of dose‐adjusted 9‐hydroxyrisperidone concentrations with CYP2D6 phenotype, P = 0.15; (c) Association of parent:metabolite ratio with CYP2D6 phenotype, P < 0.00001; (d) Association of risperidone daily dose with CYP2D6 phenotype, P = 0.02. conc, concentration; IM, intermediate metabolizer; NM, normal metabolizer; PM, poor metabolizer; UM, ultrarapid metabolizer.

CYP2D6 activity and risperidone‐associated adverse drug reactions

A large proportion of participants experienced at least one risperidone‐associated adverse drug reaction (67.8%) at some point since the initiation of therapy, and 17.3% of participants received a dose reduction (Table 2 ). The most common adverse drug reactions reported by participants were appetite change and weight gain (66% and 54% of participants, respectively). There were 24 participants (12%) who experienced hyperprolactinemia, all of whom were > 10 years of age. See Table 2 for further details on adverse drug reactions experienced by participants.

Among all adverse drug reactions, only hyperprolactinemia in participants >10 years of age was associated with CYP2D6 activity (P = 0.044, Table S10 ). The best multivariable model for hyperprolactinemia in participants > 10 years old (lowest AIC) included sex (with males having lower risk), ratio‐defined CYP2D6 activity (positively correlated with higher risk), and an interaction term for sex and ratio‐defined CYP2D6 activity (the association between ratio‐defined activity and hyperprolactinemia only occurring in females).

DISCUSSION

Using an established biomarker, we investigated the CYP2D6 enzymatic activity of two alleles more common in people historically excluded as participants from pharmacogenetic studies on risperidone metabolism. The activity scores we derived markedly differed from those reported in the literature for other CYP2D6 substrates. To our knowledge, this is the largest real‐world cohort investigating the impact of sub‐Saharan African‐specific CYP2D6 alleles on risperidone exposure.

CYP2D6 *17 and *29 alleles are predominantly found in individuals with sub‐Saharan African ancestry. ²⁹ , ³⁰ The frequencies we observed for these alleles (19% for *17 and 11% for *29) are in line with other published studies of Western Africans. ³⁰ Initial studies for *17 ¹⁸ , ¹⁹ , ³¹ , ³² , ³³ and *29 ²⁴ found reduced CYP2D6 enzyme activity associated with these alleles using CYP2D6 substrate debrisoquine across a wide range of groups with substantial sub‐Saharan African ancestry (Black American, Tanzanian, Ethiopian, Zimbabwean participants). Both alleles were eventually assigned activity values of 0.5 when scoring systems were later developed ³⁴ based on this evidence, though there was evidence for substrate specificity, particularly for codeine. In contrast, our study estimated activity scores of 2.0 and 0.0 for *17 and *29, respectively.

These observed activity value discrepancies revisit previous questions regarding the potential substrate‐specific nature of CYP2D6 function. Past studies have reported that CYP2D6 activity varies by drug substrate, with the substrate specificity being most pronounced for *17. ²⁰ , ²¹ , ³⁵ Indeed, polymorphisms in *17 cause unique amino acid changes within the active site of CYP2D6. ³⁶ This is particularly true for amino acid substitution T107I of *17, which is thought to impact substrate recognition. ³⁷ Altogether, these modifications in *17 affect both substrate binding affinity and catalytic efficiency, potentially creating substrate‐specific effects of larger magnitude than other alleles. ³⁶

To our knowledge, the only other study reporting effects of *17 on risperidone had incredibly small numbers for the specific comparisons relevant to our analysis. Particularly, Cai et al. compared the risperidone parent:metabolite ratio between *17/null (n = 3) and *1/null + *2/null (n = 10) genotypes. ³⁸ Despite the lower sample size, results in that study also showed an increase in risperidone metabolism for *17, including an effect size that agrees with our data. ³⁸ Risperidone may also be one of only two CYP2D6 substrates for which *17 is known to increase metabolism, excluding gene duplications. Haloperidol is the other. ³⁹ Numerous other substrates (sparteine, debrisoquine, metoprolol and dextromethorphan) show wide variation in the activity of this allele, but only for reduced CYP2D6 activity. ²¹ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ For instance, one study investigating four substrates showed substantially lower CYP2D6 activity when debrisoquine or dextromethorphan was used as the substrate compared with codeine or metoprolol in the presence of *17. ²¹ Additionally, studies have reported *17 having no function for some substrates (brexpiprazole and tedatioxetine). ²⁷ , ⁴⁰ Future studies using additional substrates would provide further clarity on this topic.

Our study has important clinical implications. The Association for Molecular Pathology includes CYP2D6*17 and *29 among Tier 1 (i.e., “must test”) alleles for CYP2D6 testing. ¹ CPIC provides a “clinical function” for each allele, categorizing them as having no function, decreased function, normal function, or increased function. ³⁴ For CYP2D6, CPIC uses an activity scoring system and assigns each allele an activity value. The activity values for each allele are summed to obtain an activity score, which is translated into a metabolizer phenotype. ⁴ The phenotype is then used to guide pharmacogenetic prescribing in clinical practice. However, to date, these guidelines assign activity scores without considering substrate specificity and are based on results from the limited number of drug substrates shown above. ⁴ Our results may compel CPIC to reassess CYP2D6 activity values assigned to *17 and *29 for risperidone. Furthermore, these findings may compel future investigators and pharmacogenetic guideline working groups to more carefully consider the drug substrate in question when assigning scores, especially for allele‐substrate pairs that are understudied.

The exploration of substrate specificity for understudied alleles in clinical studies would not be a trivial feat. Eurocentrically biased study populations would be less likely to have adequate allele frequencies for many of these investigations. ⁴⁴ , ⁴⁵ Participants from groups enriched with these alleles are difficult to recruit or identify in existing data resources, ⁷ especially considering the need to analyze multiple CYP2D6 substrates concurrently. Thus, another implication for our study is that it underscores the critical need for more inclusion in pharmacogenomic research. More genetic diversity in pharmacogenomics will enable the benefits of precision medicine to extend to all populations and prevent the exacerbation of preexisting health disparities.

We did not observe a strong relationship between CYP2D6 variation and adverse drug reactions in our study. Study design may have contributed to this lack of association; adverse drug reactions were reported by parents or guardians who may not have been able to precisely describe the symptoms experienced by the pediatric patients. Furthermore, they could have been subject to recall bias, as patients may have been on risperidone for many years and experienced an adverse event early in therapy, or selection bias, since those that had serious adverse events likely discontinued use of risperidone. In contrast, our biomarker of CYP2D6 activity was collected prospectively and is a more objective endpoint. This measurement is ultimately more reliable in illustrating the impact of our CYP2D6 alleles on enzyme activity. Future studies with more robust adverse drug reaction data would complement our findings.

Our study is not without limitations. First, our results were generated from the study population of a single center within Nigeria, and thus may not apply to other groups with predominant sub‐Saharan African ancestry. Nevertheless, results from Cai et al. are from a different study population of sub‐Saharan African ancestry participants (i.e., Black risperidone users in Kentucky) yet still reported a similar effect for CYP2D6*17 on risperidone, ³⁸ suggesting our findings may generalize outside of Nigerian groups. Second, we cannot rule out the potential effects of environmental factors or methodological differences on the enhanced enzyme activity of CYP2D6*17. However, evidence from aforementioned prior studies points to substrate specificity rather than these other factors. Third, recruitment for this investigation was based on risperidone use in a real‐world cohort, so it is not possible to generate data from this study population for other CYP2D6 drug substrates. A clinical trial design investigating multiple substrates in the same study population would provide even more compelling results regarding the potential of substrate specificity.

In conclusion, using the largest cohort of its kind, we found activity scores for sub‐Saharan African‐specific CYP2D6 alleles that challenge existing assumptions about CYP2D6 enzyme functionality. Importantly, these findings have the potential for improving risperidone prescribing for patients with sub‐Saharan African ancestry, a group historically excluded and currently underrepresented in pharmacogenetic studies.

FUNDING

Financial support for this work was received from the Pharmacogenomics Global Research Network (O.K., A.O.‐O., L.B.R.), the Cincinnati Children’s Center for Pediatric Genomics (L.B.R.), and the National Human Genome Research Institute (R01HG012824: A.O.‐O.).

CONFLICTS OF INTEREST

Dr. Ramsey reports receiving research funding and consultation fees from BTG Specialty Pharmaceuticals, unrelated to the current work. All other authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

O.K., S.E.V., O.A., P.T., B.R., A.O.‐O., and L.B.R. wrote the manuscript; O.K., A.O.‐O., and L.B.R. designed the research; O.K., O.A., P.T., B.R., A.O.‐O., and L.B.R. performed the research; O.K., S.V., O.A., A.O.‐O., and L.B.R. analyzed the data.

Supporting information

Data S1