Impact of CYP2D6 Metabolizer Status on Ondansetron Efficacy in Early Pregnancy Induced Nausea and Vomiting: A Case Control Study

ABSTRACT

Real‐world data investigating CYP2D6 on the efficacy of ondansetron for nausea and vomiting in pregnancy (NVP) is lacking. Evidence shows CYP2D6 ultrarapid metabolizers are at risk of ondansetron nonresponse due to increased metabolism. We conducted a retrospective cohort study of early pregnant patients on ondansetron for NVP. Genotype data for 11 CYP2D6 variants and copy number variations were translated into activity score (AS) and metabolizer status: poor (PM), intermediate (IM), normal (NM), and ultrarapid (UM) metabolizers. A total of 264 women met inclusion/exclusion criteria (99 cases and 165 controls). Multivariate regression analyses of metabolizer status and AS were adjusted for significant independent variables. The majority had gravidity of two, singleton pregnancy, a median age of 28 (interquartile range [IQR] 24–31) years, and identified as White (65.5%). Ondansetron dose was similar among the cohort and half received other antiemetics simultaneously. Clinical characteristics between cases and controls did not differ except for gestational age (8 vs. 10 weeks, p = 0.004) and primigravida rate (45.5% vs. 32.7%, p = 0.017). When adjusted for covariates, metabolizer status was not associated with response. UM/NM had non‐significantly higher risk of nonresponse (odds ratio [OR] 1.53, 95% confidence interval [CI], 0.88–2.66) compared to PM/IM. Similar trends were observed with higher CYP2D6 AS showing increased risk of nonresponse (OR 1.22, 95% CI [0.81–1.85]). This study found no significant differences in ondansetron response in early pregnancy based on CYP2D6 UM/NM versus PM/IM and AS. Additional prospective investigations are necessary to confirm the CYP2D6 effect on ondansetron efficacy in pregnant patients.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Nausea and vomiting in pregnancy (NVP) affects up to 80% of pregnant individuals and is frequently treated with ondansetron. Ondansetron metabolism is influenced partly by CYP2D6, and adult non‐pregnant studies demonstrate reduced efficacy in CYP2D6 ultrarapid metabolizers, prompting CPIC guideline recommendations to avoid ondansetron in this group. However, real‐world pharmacogenetic evidence in pregnant patients is extremely limited, and pregnancy‐induced changes in CYP2D6 further complicate the prediction of drug response.
• What question did this study address?

  • ○
    This retrospective case–control study investigated whether CYP2D6 metabolizer status or activity score predicts clinical response to oral ondansetron for NVP during early pregnancy (within 20 weeks of conception).
• What does this study add to our knowledge?

  • ○
    To date, this is the largest pharmacogenetic study examining CYP2D6 variation and ondansetron response in pregnant patients with NVP. Contrary to the hypothesis, no statistically significant association was found between CYP2D6 metabolizer status (UM/NM vs. PM/IM) or activity score and ondansetron response (OR 1.53 [0.88–2.66], p = 0.14). However, a non‐significant trend suggested increased nonresponse among ultrarapid/normal metabolizers. Clinical factors (i.e., primigravida status, earlier gestational age at exposure, and self‐identified Black race) were associated with higher risk of nonresponse.
• How might this change clinical pharmacology or translational science?

  • ○
    Findings suggest that current CYP2D6‐based ondansetron recommendations derived from non‐pregnant adults may not directly translate to early pregnancy. The results highlight the need for pregnancy‐specific pharmacogenomic frameworks, consideration of pregnancy‐related physiologic changes in multiple hepatic enzymes, and prospective studies integrating multiple metabolic pathways. Future research should aim to include larger studies enriched with CYP2D6 ultrarapid and poor metabolizers, prospective designs, and investigation of other genetic factors that may influence ondansetron response during pregnancy.

1. Introduction

Nausea and vomiting in pregnancy (NVP) is a common condition that affects up to 80% of pregnant women [1]. NVP generally starts early in the first trimester, peaking in severity between 9 and 16 weeks, with resolution occurring around 20 weeks of gestation [1]. There is a wide variation of severity and duration that can occur throughout the pregnancy. Once NV progresses, it can become more difficult to control symptoms and thus early and effective interventions are crucial. Severe and persistent NVP or hyperemesis gravidarum (HG) can lead to maternal and fetal morbidity if left untreated and is also the most common cause of hospitalization during the first half of pregnancy [1].

Ondansetron and other serotonin receptor antagonists are one of the most effective antiemetics for NVP and HG, yet treatment response varies widely among patients [2]. For NVP, recommended oral doses range from 4 to 8 mg every 8 h as needed, with intravenous administration reserved for severe and/or refractory cases. Oral ondansetron is primarily metabolized in the liver via cytochrome P450 (CYP) enzymes, mainly CYP3A4, CYP1A2, and CYP2D6. Among these, CYP2D6 is estimated to contribute to around 30% of the metabolism and ondansetron metabolites do not contribute to its antiemetic effects [3]. Although CYP3A4 and CYP1A2 contribute to its biotransformation, genetic studies have largely focused on CYP2D6 due to its well‐characterized polymorphisms and association with variability in ondansetron response [4, 5]. CYP2D6 is a highly polymorphic gene with over 170 distinct haplotypes, and copy number variations (CNVs) including gene deletions, duplications, hybrid genes with CYP2D7, and combinations thereof that differentially affect enzyme activity [6]. Currently, individuals are classified into four metabolizer status based on their CYP2D6 genotype‐based activity scores (AS). Activity ranges widely among individuals from poor metabolizers (due to deletions or two non‐functioning alleles) to ultrarapid metabolizers (due to CNV or duplications of functional alleles) [6]. Particularly, current literature shows that CYP2D6 ultrarapid metabolizers have a decreased response to ondansetron for postoperative NV in adults due to increased metabolism [4]. Thus, the Clinical Pharmacogenomics Implementation Consortium (CPIC) guideline recommends alternative therapy to ondansetron for CYP2D6 ultrarapid metabolizers [4].

Despite the frequent use of ondansetron during pregnancy, there is limited research conducted in this population. This is largely driven by the historic exclusion of pregnant women in traditional clinical trials and necessitates further investigations to tailor therapy and recommendations to this population [7, 8, 9, 10]. In addition, the increase in CYP2D6 enzyme activity throughout pregnancy can potentially affect CYP2D6 normal metabolizers to present with higher than expected CYP2D6 activity [11, 12, 13]. Clinical databanks linked to electronic health records are an invaluable resource to study medication use in this historically excluded population. Therefore, given the availability of actionable CYP2D6 genotyping and its known impact on ondansetron pharmacokinetics, our study leverages our institution's biobank to test the hypothesis that response to ondansetron in early pregnancy is influenced by CYP2D6 variation.

2. Methods and Materials

2.1. Study Population

We conducted a retrospective case–control study using Vanderbilt University Medical Center's (VUMC) biobank, BioVU, a genomic repository linked to de‐identified electronic health records (EHR) [14, 15]. BioVU accrues DNA from blood samples obtained during routine clinical care from patients who have consented to have a DNA sample collected. DNA is extracted from consented samples that would otherwise be discarded. This study was reviewed and approved by the VUMC Institutional Review Board and determined to be non‐human subject research.

Medical records from pregnant women who received ondansetron within 20 weeks of conception as part of their medical care at VUMC were manually reviewed for inclusion from January 2010 to December 2020. Women were included if they met inclusion criteria: (1) DNA sample available in BioVU, (2) live birth confirmed by diagnosis or procedure codes (Table S1), (3) pregnancy at 18 ± 1 years of age or older (dates are shifted −1 year in BioVU), (4) evidence that oral ondansetron was taken within the first 20 weeks of pregnancy, (5) have a diagnosis of NVP (Table S1) at ondansetron intake and, (6) have a follow‐up clinical encounter (i.e., phone call, outpatient or emergency room [ER] visit) after ondansetron was prescribed to assess response. Additional inclusion criteria were the availability of non‐compromised DNA. To facilitate manual chart review by creating an initial cohort, a phenotype algorithm was developed to identify pregnant women who fulfilled inclusion criteria 1–4 and all inclusion criteria were then examined during manual chart review.

We excluded women who met exclusion criteria: (1) nausea and vomiting attributable to a concurrent medical condition (i.e., active cancer or treatment for cancer, kidney stones, a flare of inflammatory bowel disease, sickle cell crisis), (2) substance use (e.g., marijuana, cocaine, opioid use) documented at time of ondansetron exposure, or (3) had indeterminate CYP2D6 phenotype. In the case of a patient with multiple pregnancies, only the first pregnancy that met inclusion and exclusion criteria was included.

2.2. Medication Exposure and Clinical Covariate Definitions

Oral ondansetron (dose and frequency) and other medications (i.e., pyridoxine, promethazine, metoclopramide, chlorpromazine, and concomitant CYP2D6 inhibitors) along with relevant covariates including demographic data and clinical features were extracted through manual chart review. To select only women exposed to oral ondansetron within the first 20 weeks of pregnancy, we defined an exposure window of 20 ± 4 weeks using the date of conception plus 24 weeks. The date of conception was calculated using the gestational age at delivery or, if not recorded, the gestational age at the earliest prenatal care visit.

2.3. Case–Control Definitions

Nonresponse to ondansetron for nausea with or without vomiting within 20 weeks of gestation after a prescription for oral ondansetron was defined as any of the following: (1) ER visit, infusion center, and/or hospital admission due to nausea with or without vomiting, (2) received intravenous ondansetron, (3) prescribed other medications for nausea (Table S2). Response to ondansetron was defined as a lack of any of the above findings. Responders (cases) and non‐responders (controls) were mutually exclusive and defined through chart review. Two reviewers independently reviewed all charts. In cases of discordant adjudication status or complex calls, a third independent review was added, and final determination was based on discussion of all three reviewers.

2.4. CYP2D6 Genotyping

Following the Association for Molecular Pathology recommendations for clinical testing, we analyzed DNA samples for 11 CYP2D6 single nucleotide variants (rs16947, rs1080985, rs35742686, rs3892097, rs5030655, rs5030867, rs5030656, rs1065852, rs28371706, rs59421388, and rs28371725) and three assays for CNV targeting intron 6, exon 6 and exon 9, to define CYP2D6 functional status and AS [6]. DNA samples were genotyped using TaqMan assays (Thermo Fisher Scientific, Walham, MA) in the Vanderbilt Technologies for Advanced Genomics (VANTAGE) Core Laboratory according to manufacturers' protocols. All samples with ≥ 95% call rates and passing quality control were included. CYP2D6 diplotype calls from genotype data were performed informatically and subsequently manually reviewed. CYP2D6 genotype and phenotype assignments were according to CPIC recommendations as well as PharmVar's CYP2D6 structural variation guidance [16, 17]. First, diplotypes were translated to AS and subsequently into phenotype or metabolizer status; poor metabolizers (PMs, AS = 0) intermediate metabolizers (IMs, AS = 0.25–1), normal metabolizers (NMs, AS = 1.25–2.25), and ultrarapid metabolizers (UMs, AS > 2.25) [4, 18]. For any individual with a CNV of three or more gene copies, the AS may result in a range and ambiguous metabolizer status prediction given that testing did not determine which allele was duplicated. These individuals with ambiguous AS were removed from subsequent analyses. Further, samples with discordant CNV results and CYP2D6‐2D7 or CYP2D7‐2D6 hybrid structures were included in the analysis if their metabolizer status could be unambiguously determined (Table S3).

2.5. Statistical Analysis

Categorical variables are described as frequency and percentages (%) and continuous variables as median and interquartile ranges (IQR). Clinical characteristics, CYP2D6 genotypes, metabolizer status, and AS were compared in cases and controls using Chi‐square or Kruskal Wallis tests. For primary analyses, we performed a logistic regression to define the role of (a) genetically predicted metabolizer status as a binary variable (UM/NM vs. PM/IM), and (b) CYP2D6 AS on ondansetron response. For all analyses, we accounted for the effect of phenoconversion due to potential exposure to strong CYP2D6 inhibitors (Table S4) [18]. The effect of metabolizer status and AS on NVP is shown as odds ratio (95% confidence interval) – OR (95% CI). Post hoc analysis included analysis of UM/NM with AS ≥ 2 or at least two functional alleles vs. PM/IM with AS ≤ 1. Regression analyses were adjusted for variables identified as significant in univariate analyses. A two‐sided p ≤ 0.05 was considered statistically significant for all tests. Hardy–Weinberg equilibrium (p < 0.001) using the exact test was assessed and adjusted by EHR reported race. Statistical analyses were performed using Rstudio (version 4.1) and STATA software program Version 17 (STATA, TX).

3. Results

3.1. Characteristics of Cohort

A total of 321 patients qualified as a case or control, while 264 patients met all inclusion and exclusion criteria (99 cases of nonresponse (Table S5) and 165 controls). The median gravidity of the cohort was 2 IQR (1–3) and 95% had a singleton pregnancy (Table 1). The median age was 28 (24–31) years old, 66% identified as White and 91% as non‐Hispanic. Median ondansetron dose was similar among the cases and controls (16 [12–24] mg/day) and half of the cohort received other antiemetic agents (pyridoxine, promethazine, metoclopramide, or chlorpromazine) simultaneously with ondansetron. Cases and controls demonstrated a significant difference in gestational age (8 vs. 10 weeks, p = 0.004) and primigravida rates (45% vs. 33%, p = 0.017). There were more self‐identified Black patients among cases than controls (33% vs. 19%, p = 0.017, Table 1). Among cases, the median time between first mention of ondansetron prescription in the EHR and fulfilling the definition of non‐response was 11 (5–18) days (Table S5).

TABLE 1.

Clinical characteristics of the selected population.

Patient baseline characteristics	Total cohort a (N = 264)	Case (N = 99)	Control (N = 165)	p
Age, median (IQR), years	27.50 (24–31)	27.00 (23–30.5)	28.00 (24–31)	0.223
Pre‐pregnancy BMI, median (IQR), kg/m2	25.00 (21–30)	24.00 (21–28.5)	25.00 (22–31.5)	0.124
< 18.5	13 (4.9%)	7 (7.1%)	6 (3.6%)	0.425
18.5–24.9	89 (33.7%)	37 (37.4%)	52 (31.5%)	0.662
25.0–29.9	53 (20.1%)	20 (20.2%)	33 (20.0%)	0.884
≥ 30.0	55 (20.8%)	19 (19.2%)	36 (21.8%)	0.473
Smoking during pregnancy (%)	18 (6.8%)	3 (3.0%)	17 (10.3%)	0.117
Pregnancy factors
Gestational age, median (IQR), weeks	9 (7–12)	8 (7–11)	10 (8–13)	0.004
Singleton (%)	250 (94.7%)	94 (95.0%)	156 (94.5%)	1
First pregnancy (%)	99 (37.5%)	45 (45.5%)	54 (32.7%)	0.017
EHR‐identified race
White	173 (65.5%)	59 (59.6%)	114 (69.0%)	0.151
Black	65 (24.6%)	33 (33.3%)	32 (19.4%)	0.017
Other	12 (4.5%)	4 (4.0%)	8 (4.8%)	1
Unknown	14 (5.3%)	3 (3.0%)	11 (6.7%)	0.321
EHR‐identified ethnicity
Non‐Hispanic	239 (90.5%)	91 (91.9%)	148 (89.7%)	0.704
Hispanic	23 (8.7%)	8 (8.1%)	15 (9.1%)	0.955
Unknown	2 (0.8%)	0	2 (1.2%)	0.714
CYP2D6 phenotype b
Ultrarapid metabolizer	12 (4.5%)	6 (6.1%)	6 (3.6%)	0.542
Normal metabolizer	147 (55.7%)	57 (57.6%)	87 (52.7%)	0.523
Intermediate metabolizer	95 (36.0%)	35 (35.4%)	60 (36.4%)	0.974
Poor metabolizer	10 (3.8%)	1 (1.0%)	9 (5.5%)	0.134
Initial ondansetron dose, mg/d	16 (12–24)	16 (12–24)	16 (12–24)	0.446
Concomitant medication use
Vitamin B6	48 (18.2%)	14 (14.1%)	34 (20.6%)	0.249
Other c	84 (31.8%)	38 (38.4%)	46 (27.8%)	0.102

^a Pre‐pregnancy BMI was available for 210 subjects in the total cohort.

^b Poor metabolizers (PMs, AS = 0) intermediate metabolizers (IMs, AS = 0.25–1), normal metabolizers (NMs, AS = 1.25–2.25), and ultrarapid metabolizers (UMs, AS > 2.25).

^c Included doxylamine and pyridoxine (Diclegis and Bonjesta) combination agents as well as promethazine, metoclopramide, chlorpromazine, among other types of antiemetic medications.

3.2. CYP2D6 Phenotypes

Of the 264 genotyped individuals, phenotype was assigned as: 12 UMs (4.5%), 147 NMs (55.7%), 95 IMs (36%), and 10 PMs (3.8%). Six patients had AS ranging from IM to NM, while one patient had AS ranging from NM to UM. Using a conservative approach, we assigned these seven patients as IM and UM for analysis, respectively. Nine individuals (five IMs and four NMs) were exposed to a strong CYP2D6 inhibitor and were assigned as CYP2D6 PMs (Table S4). Cases and controls did not differ in the metabolizer status rates; however, there were non‐significant higher rates of UMs (6.1% vs. 3.6%) and lower rates of PMs (1% vs. 5.5%) in cases compared to the control group.

The logistic regression adjusted for primigravida status, gestational age at ondansetron exposure, and reported race did not show a significant difference in ondansetron response when comparing UM/NM to PM/IM or AS (UM/NM (AS ≥ 2) compared to PM/IM (AS ≤ 1), Figure 1). However, individuals in the UM/NM group exhibited a nonsignificant trend toward higher odds of nonresponse compared to PM/IM (OR 1.53 [0.88–2.66], p = 0.14). Analysis using CYP2D6 AS showed similar trends of increased risk of nonresponse to ondansetron with higher AS (OR 1.22, 0.81–1.85, p = 0.35). Post hoc analysis of the UM/NM group that had AS ≥ 2 showed a non‐significant interaction (OR 1.22, 0.89–1.67, p = 0.54). Primigravida status, earlier gestational age at ondansetron exposure, and reported Black race were associated with increased risk of nonresponse to ondansetron in all analyses.

FIGURE 1.

Forest plot depicting the risk for ondansetron treatment failure with CYP2D6 activity score. CI, confidence interval; CYP2D6 activity score as a continuous variable and accounts for phenoconversion.

4. Discussion

This is the first study that investigates CYP2D6 variability and clinical response to ondansetron for nausea and/or vomiting during early pregnancy. Contrary to our hypothesis, we did not find a statistically significant association between genetically predicted CYP2D6 metabolizer status or AS and ondansetron response. There was a trend in the direction of effect that suggests an increased nonresponse among CYP2D6 UM/NM and those with higher AS, while PM/IM and lower AS may be associated with reduced risk of nonresponse. Additionally, primigravida status and earlier gestational age were associated with nonresponse. Cases were more likely than controls to identify as Black, which is surprising considering that previous studies suggest lower CYP2D6 activity in Black Americans [19]. There are limited studies that investigate the impact of CYP2D6 variation in pregnant women on ondansetron. One study of 62 pregnant women in the perioperative period during cesarean section who received ondansetron 4–8 mg IV every 6 h investigated associations with QT prolongation. The study observed 14 pregnant patients with QTc prolongation (median shift = 5 ms, p = 0.02) but it was not associated with CYP2D6 variants or AS; however, the route of administration for ondansetron was IV opposed to oral which may minimize the effects of CYP2D6 metabolizer status [20].

Much is not understood about the effect of pregnancy and induction of the CYP2D6 gene activity, which could affect all CYP2D6 phenotypes except CYP2D6 PMs. Although there is a trend in the available literature that shows CYP2D6 substrate drugs have increased metabolism and decreased concentrations for normal and ultrarapid metabolizers in pregnancy compared to postpartum [21]. One patient case report suggests a dynamic induction of CYP2D6 enzymes as pregnancy progresses, with early pregnancy (14 weeks gestation) representing the lowest level of CYP2D6 induction compared to mid and late‐pregnancy time points [11]. We included only early pregnancy patients, which could minimize CYP2D6 induction effects from pregnancy changes. Although adult studies in non‐pregnant populations show ondansetron response affected by CYP2D6 variation, more studies are needed, especially ones enriched with CYP2D6 UMs and potentially PMs [7, 10, 22].

This study focused on CYP2D6 pharmacogenetics due to its well‐characterized polymorphisms and established clinical relevance in ondansetron response. However, we recognize that pregnancy induces physiologic changes in multiple hepatic enzymes, including CYP3A4 and CYP1A2, which also contribute to ondansetron metabolism. Future studies with larger sample sizes and broader genotyping could explore the combined impact of these metabolic pathways to better inform personalized antiemetic therapy in pregnancy.

4.1. Strengths and Limitations

A strength of our study is that this is the largest pharmacogenetic study to date of pregnant patients with NVP on ondansetron; however, given the nature of our population's CYP2D6 frequency distributions, a larger sample enriched with CYP2D6 UMs would increase the generalizability of our finding. Secondly, we included three different target assays for CYP2D6 CNV. Given the complex structural arrangements of the CYP2D6 gene and possible pseudogenes, this increases our accuracy of interpreting CYP2D6 genotypes and therefore calling phenotypes.

An important limitation is the retrospective nature of our study in which we cannot control for confounders such as adherence to ondansetron, missing data (e.g., unreported over‐the‐counter therapies, duration of therapy) or identifying controls that were lost to follow‐up at our institution. However, we only selected individuals who had enough information to assess intake and response through manual chart review. The use of EHR data also decreases the risk of recall bias since exposure data is obtained from clinical encounters. Prospective clinical trials in pregnant populations are often limited by ethical and logistical challenges, making retrospective case–control designs a practical and efficient alternative.

Several other genes may influence ondansetron response; however, interpretation of genotype to phenotype and how genetic variation affects response is less established; hence they were outside the scope of this study. Future research is needed to evaluate how ABCB1 (encodes P‐glycoprotein, a drug efflux transporter), SLC22A1 (OCT11, an organic cation transporter), HTR3A (5‐HT₃A receptor subunit), CYP1A2, and CYP3A4 (hepatic enzymes) can affect ondansetron response. Lastly, cases and controls were not matched, which would require a larger sample size.

5. Conclusion

In this cohort case control study of 264 early pregnant patient population, ondansetron response was not significantly associated with CYP2D6 metabolizer status (UM/NM vs. PM/IM) or AS. Future prospective studies are necessary to further evaluate this association given our findings.

Author Contributions

M.L., M.M.S., and V.K.K. wrote the paper. M.M.S., D.R.V.E., and V.K.K. designed the research. M.L., M.M.S., S.E.H., D.F.F., C.G., J.K., M.S., S.D.S., H.H.O., and V.K.K. performed the research. J.J. performed the analysis. M.L., S.G., A.G., and W.‐Q.W. analyzed the data.

Funding

This work was supported by NIH P50 HD106446 and the Vanderbilt Integrated Center of Excellence in Maternal and Pediatric Precision Therapeutics (VICE‐MPRINT). The project publication described was also supported by CTSA award No. UL1 TR002243 from the National Center for Advancing Translational Sciences. Its contents are solely the responsibility of the authors and do not necessarily represent official views of the National Center for Advancing Translational Sciences or the National Institutes of Health. M.M.S. was supported for this work by K12AR084232. J.K. was supported for this work by TL1TR002244.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: cts70523‐sup‐0001‐DataS1.docx.