Propensity score stratified level 1 refers to analyses adjusted for treatment indications and associated factors. Propensity score stratified level 2 refers to analyses adjusted for all potential confounding variables as listed in eTable 3 in the Supplement. High-dimensional propensity score stratified refers to analyses adjusted for 200 empirically identified covariates, in addition to the prespecified covariates.
Sensitivity analyses include changing the reference group to women exposed to any of the 3 other antiemetics; filling at least 2 ondansetron prescriptions; changing the exposure window to 6 to 12 weeks after last menstrual period; using pregnancies with first exposure to ondansetron 5 to 8 months after last menses (reference); and changing the exposure window to 5 to 8 months after last menses (negative control analysis).
eTable 1. Available evidence on ondansetron and risk of malformations
eTable 2. Definition of congenital malformations
eTable 3. Cohort characteristics of pregnancies with and without exposure to ondansetron during the first trimester
eTable 4. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between ondansetron-exposed and unexposed pregnancies - Different levels of adjustment
eTable 5. Absolute risk and adjusted risk difference of congenital malformations among infants born to mothers with and without exposure to ondansetron
eTable 6. Cohort characteristics of pregnancies exposed to ondansetron and pregnancies exposed to other anti-emetics (pyridoxine, promethazine, or metoclopramide) during the first trimester
eTable 7. Cohort characteristics of pregnancies exposed to ondansetron and pregnancies exposed to metoclopramide during the first trimester
eTable 8. Cohort characteristics of pregnancies exposed to ondansetron and pregnancies exposed to promethazine during the first trimester
eTable 9. Cohort characteristics of pregnancies exposed to ondansetron and pregnancies exposed to pyridoxine during the first trimester
eTable 10. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between pregnancies exposed to ondansetron and pregnancies exposed to other anti-emetics (pyridoxine, promethazine, or metoclopramide)
eTable 11. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between pregnancies exposed to ondansetron and pregnancies exposed to metoclopramide
eTable 12. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between pregnancies exposed to ondansetron and pregnancies exposed to promethazine
eTable 13. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between pregnancies exposed to ondansetron and pregnancies exposed to pyridoxine
eTable 14. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between ondansetron-exposed and unexposed pregnancies - Sensitivity Analyses: ≥2 prescriptions
eTable 15. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between ondansetron-exposed and unexposed pregnancies - Sensitivity Analyses: prescription during 6-12 weeks
eTable 16. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between ondansetron-exposed (prescription during 5-8 months) and unexposed pregnancies - Sensitivity Analyses: negative control analysis
eTable 17. Relative risks and 95% confidence intervals comparing the risk for congenital malformations between pregnancies with ondansetron exposure during the first trimester and pregnancies with the first ondansetron exposure during gestational months 5 to 8 - Sensitivity Analyses: ondansetron T1 vs 5-8 months
eTable 18. Probability of live birth
eFigure 1. Ondansetron use during the first trimester of pregnancy over time
eFigure 2. Corrected relative risk for the association between ondansetron use during the first trimester and any congenital malformation, starting from an observed RR of 1.01
eFigure 3. Corrected relative risk for the association between ondansetron use during the first trimester and cardiac malformations, starting from an observed RR of 0.99.
eFigure 4. Corrected relative risk for the association between ondansetron use during the first trimester and oral clefts, starting from an observed RR of 1.24.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Huybrechts KF, Hernández-Díaz S, Straub L, et al. Association of Maternal First-Trimester Ondansetron Use With Cardiac Malformations and Oral Clefts in Offspring. JAMA. 2018;320(23):2429–2437. doi:10.1001/jama.2018.18307
Is the use of ondansetron during pregnancy associated with increased risk of cardiac malformations and oral clefts in offspring?
In this cohort study including 1.8 million pregnancies, first trimester ondansetron use was associated with an increased risk of oral clefts (3 additional cases per 10 000 women treated; adjusted relative risk, 1.24) but not cardiac malformations.
Maternal use of ondansetron in the first trimester was associated with a small increased risk of oral clefts but no increased risk of cardiac malformations.
Evidence for the fetal safety of ondansetron, a 5-HT3 receptor antagonist that is commonly prescribed for nausea and vomiting during pregnancy, is limited and conflicting.
To evaluate the association between ondansetron exposure during pregnancy and risk of congenital malformations.
Design, Setting, and Participants
A retrospective cohort study nested in the 2000-2013 nationwide Medicaid Analytic eXtract. The cohort consisted of 1 816 414 pregnancies contributed by 1 502 895 women enrolled in Medicaid from 3 months before the last menstrual period through 1 month or longer after delivery; infants were enrolled in Medicaid for at least 3 months after birth. The final date of follow-up was December 31, 2013. Analyses were conducted between November 1, 2017, and June 30, 2018. Propensity score stratification was used to control for treatment indication and other confounders.
Ondansetron dispensing during the first trimester, the period of organogenesis.
Main Outcomes and Measures
Primary outcomes were cardiac malformations and oral clefts diagnosed during the first 90 days after delivery. Secondary outcomes included congenital malformations overall and subgroups of cardiac malformations and oral clefts.
Among 1 816 414 pregnancies (mean age of mothers, 24.3 [5.8] years), 88 467 (4.9%) were exposed to ondansetron during the first trimester. Overall, 14 577 of 1 727 947 unexposed and 835 of 88 467 exposed infants were diagnosed with a cardiac malformation, for an absolute risk of 84.4 (95% CI, 83.0 to 85.7) and 94.4 (95% CI, 88.0 to 100.8) per 10 000 births respectively. The absolute risk of oral clefts was 11.1 per 10 000 births (95% CI, 10.6 to 11.6; 1921 unexposed infants) and was 14.0 per 10 000 births (95% CI, 11.6 to 16.5; 124 exposed infants). The risk of any congenital malformation was 313.5 per 10 000 births (95% CI, 310.9 to 316.1; 54 174 unexposed infants) and was 370.4 (95% CI, 358.0 to 382.9; 3277 exposed infants). The adjusted relative risk (RR) for cardiac malformations was 0.99 (95% CI, 0.93 to 1.06) and the adjusted risk difference (RD) was −0.8 (95% CI, −7.3 to 5.7 per 10 000 births). For oral clefts, the adjusted RR was 1.24 (95% CI, 1.03 to 1.48) and the RD was 2.7 (95% CI, 0.2 to 5.2 per 10 000 births). The adjusted estimate for congenital malformations overall was an RR of 1.01 (95% CI, 0.98 to 1.05) and an RD of 5.4 (95% CI, −7.3 to 18.2 per 10 000 births).
Conclusions and Relevance
Among offspring of mothers enrolled in Medicaid, first-trimester exposure to ondansetron was not associated with cardiac malformations or congenital malformations overall after accounting for measured confounders but was associated with a small increased risk of oral clefts.
Quiz Ref IDNausea and vomiting occur in up to 80% of pregnancies.1 About one-third of women with nausea and vomiting during pregnancy have symptoms that are clinically significant.1 A small percentage will develop hyperemesis gravidarum, which is the most common indication for hospital admission during the first part of pregnancy.1 Early treatment of nausea and vomiting during pregnancy is recommended to provide symptom relief and prevent progression to hyperemesis gravidarum.1 Although not formally approved for the treatment of nausea and vomiting during pregnancy, ondansetron, a 5-HT3 receptor antagonist, has rapidly become the most frequently prescribed drug for nausea and vomiting during pregnancy in the United States because of its perceived superior antiemetic effects and improved adverse-effect profile compared with treatment alternatives.2-5 In 2014, an estimated 22% of pregnant women used ondansetron in the United States.6
Nausea and vomiting during pregnancy typically occurs during the first trimester, the most sensitive time for exposure to teratogens because of organogenesis. The available evidence on the fetal safety of ondansetron is limited and conflicting (eTable 1 in the Supplement). While some studies suggested no increase in birth defects in women who took ondansetron early in pregnancy,7-9 others suggested the possibility of a doubling of the risk of cleft palate and cardiac malformations.10-12 Conflicting data leave pregnant women and clinicians unsure with respect to the appropriate risk-benefit trade-off for ondansetron use. The American College of Obstetricians and Gynecologists recently concluded that there are insufficient data on fetal safety with ondansetron exposure and further studies are warranted.13
The objective of this study was to investigate the association between exposure to ondansetron during the first trimester of pregnancy and risk of congenital malformations in offspring, focusing on cardiac malformations and oral clefts, in a national cohort of publicly insured pregnant women.
The research was approved by the institutional review board of Brigham and Women’s Hospital, which granted a waiver of informed consent.
We used the pregnancy cohort nested in the nationwide Medicaid Analytic eXtract (MAX) for the years 2000 through 2013 (the most recent data available at the time of study conduct, which was conducted between November 1, 2017, and June 30, 2018), the development of which has been described elsewhere14 and which has been used extensively to study the safety of medications during pregnancy.15-18 MAX data include demographic and insurance enrollment information, medical visits and hospitalizations, diagnoses and procedures received as an inpatient or an outpatient, and prescriptions filled on an outpatient basis. Quiz Ref IDWomen aged 12 through 55 years were required to have Medicaid coverage from 3 months before the date of the last menstrual period to 1 month after delivery. Infants were required to have coverage through Medicaid for the first 3 months of life unless they died sooner. These restrictions did not affect the age or race distribution but resulted in a decrease in the proportion of low-income women eligible through the occurrence of pregnancy and a corresponding increase in the proportion eligible independently of pregnancy.16 Pregnancies with exposure to a known teratogenic medication (ie, warfarin, antineoplastic agents, lithium, isotretinoin, misoprostol, thalidomide) during the first trimester (n = 3562) and pregnancies with a chromosomal abnormality (n = 3156) were excluded.
Women were considered exposed if they filled at least 1 ondansetron prescription during the first 3 months of pregnancy. Women who did not fill a prescription for ondansetron during the 3 months before the start of pregnancy through the end of the first trimester were considered unexposed. Women who filled an ondansetron prescription during the 3 months before the start of pregnancy were excluded to avoid contaminating the reference group with women who still had medication available for ingestion after the start of pregnancy. Women who filled a prescription during the first 90 days of pregnancy for pyridoxine (with or without doxylamine), promethazine, metoclopramide, or any of these alternative pharmacological treatments were used as alternative reference groups in additional analyses. Although these medications are available over-the-counter, Medicaid recipients frequently obtain prescriptions, allowing their use to be observed in our data and making this analysis possible.19
The presence of congenital malformations was defined using algorithms based on inpatient or outpatient diagnoses and procedure codes in the maternal (first month after delivery) or infant (first 3 months after date of birth) record, which have been shown to identify congenital malformations with high specificity.20,21 The maternal record was considered because Medicaid claims are sometimes recorded under the mother before the infant’s eligibility has been processed.22 Motivated by findings from earlier studies suggesting a possible increased risk, the primary outcomes were defined as cardiac malformations and oral clefts. In secondary analyses, specific subgroups of cardiac malformations and oral clefts (ie, palate, lip, or lip and palate) were evaluated along with congenital malformations overall. In exploratory analyses, the risks of other specific malformations were evaluated (eTable 2 in the Supplement for the definition of all malformation outcomes).
A broad range of potential confounders and proxies for confounders were considered, including treatment indication (nausea and vomiting during pregnancy, hyperemesis gravidarum) and associated conditions (weight loss, electrolyte and laboratory abnormalities, dehydration, gastroesophageal reflux), calendar year, state of residence, age, race, multiple gestation, maternal conditions, concomitant medication use, and general markers of the burden of illness. Race or ethnic group was considered to represent a potentially important confounding factor and was categorized based on information submitted to the Centers for Medicare & Medicaid Services by individual states, which was based on information that had been collected and coded from Medicaid applications. Maternal morbidity and concomitant medication use were measured from 3 months before the start of pregnancy to the end of the first trimester. General markers of the burden of illness were assessed during the 3 months before but not during pregnancy to avoid these measures being affected by early awareness of possible pregnancy complications. Maternal conditions were assessed based on diagnostic codes and included psychiatric and neurological conditions (anxiety, depression, migraine, or other headache) and other chronic conditions (diabetes, hypertension, renal disease, Crohn disease, irritable bowel syndrome, ulcerative colitis, overweight or obesity, underweight, illicit drug or alcohol abuse or dependence, smoking). Concomitant medications assessed included psychotropic medications (anticonvulsants, antidepressants, benzodiazepines), triptans, oral hypoglycemics, insulin, antihypertensives, progestins, corticosteroids, and suspected teratogens (fluconazole, methimazole, danazol, propylthiouracil, aminoglycosides, folic acid antagonists, potassium iodide, tetracycline). General markers of comorbid illness and disease severity included the obstetric comorbidity index,23 number of distinct prescriptions for medications other than antiemetics, number of distinct diagnoses, number of outpatient visits, hospitalizations (yes/no, number, days), and emergency department visits.24
Balance with respect to baseline characteristics between women treated with ondansetron and the untreated reference group was assessed using standardized differences. An absolute standardized difference greater than 0.1 was considered an indicator for substantial imbalances between the 2 exposure groups.25 Absolute risks for each of the outcomes and unadjusted relative risks and risk differences with their 95% confidence intervals were calculated. Prevalence at birth was used as a proxy for absolute risks under the assumption that only a small proportion of fetuses with nonsyndromic defects would die or be terminated in utero. The actual risks may therefore be slightly larger. In the first adjusted analysis, we accounted for the possible confounding effect of the indication for treatment and its associated factors (nausea and vomiting during pregnancy, hyperemesis gravidarum, and the use of other antiemetics). In the second adjusted analysis, we accounted for all potential confounding variables described above (eTable 3 in the Supplement for a comprehensive list of covariates included). Propensity scores, estimated using logistic regression, were used to efficiently adjust for a large number of covariates. After trimming observations from the nonoverlapping regions of the propensity score distribution, we created 50 equally sized propensity score strata based on the distribution among the treated women.26 In the outcome models, the untreated observations were weighted using the distribution of the treated among propensity score strata. In confirmatory analyses, high–dimensional propensity score analyses that included 200 empirically defined covariates, in addition to the prespecified covariates, were conducted to account for potential residual confounding.27 Adjusted relative risks and risk differences were estimated using a log-binomial model.
Sensitivity analyses were conducted to test the robustness of the primary results. The reference group was changed to women who filled a prescription for a different antiemetic medication during the first trimester because they might be more comparable with women exposed to ondansetron than women who were never treated with antiemetic agents during pregnancy. To evaluate the potential effect of exposure misclassification, exposure was redefined as having filled 2 or more prescriptions for ondansetron during the first trimester. Because the period of greatest sensitivity to teratogens for oral clefts is likely the later part of the first trimester, we redefined the exposure window as 6 to 12 weeks after the date of last menstrual period. In a negative control analysis, we assessed the risk of malformations in women who filled their first ondansetron prescription in gestational months 5 to 8, which is after the etiologically relevant window. A null finding in this analysis provides indirect evidence of no substantial residual confounding. We also used this group with first ondansetron exposure in months 5 to 8 as an alternative reference group. Because the cohort included live births only, the potential effect of differences in the proportion of pregnancy losses among women treated with ondansetron vs those untreated within levels of covariates used in the adjustment were conducted, as previously described (eAppendix in the Supplement: Quantifying the potential effect of restriction to livebirths).15
All analyses were conducted using SAS version 9.4 (SAS Institute Inc). Precision around the estimates of absolute risk and the measures of association are provided using 2-sided 95% confidence intervals. No adjustments were made for multiple comparisons, and results of secondary analyses should be interpreted as exploratory.28-30
The cohort consisted of 1 816 414 pregnancies, of which 88 467 (4.9%) were exposed to ondansetron during the first trimester. Use increased from 0.01% in 2000 to 12% in 2013 (eFigure 1 in the Supplement). Most women (1 251 216; 83.3%) contributed 1 pregnancy to the cohort, 203 037 (13.5%) contributed 2 pregnancies, and 48 642 (3.2%) contributed 3 or more pregnancies. Exposed women were more likely to be white, to have a diagnosis of psychiatric and neurological conditions, and to smoke. They were also more likely to fill a prescription for other medications used to treat nausea and vomiting during pregnancy (metoclopramide, promethazine, pyridoxine), for psychotropic medications, for corticosteroids, and for suspected teratogens. Most markers of comorbid illness and disease severity were also elevated among ondansetron-exposed compared with unexposed women (Table 1 and eTable 3 in the Supplement).
The risk of cardiac malformations was 94.4 (95% CI, 88.0-100.8) per 10 000 among ondansetron-exposed pregnancies and 84.4 (95% CI, 83.0-85.7) per 10 000 among unexposed pregnancies (estimated based on 835 exposed and 14 577 unexposed cases), for an unadjusted relative risk (RR) of 1.12 (95% CI, 1.04-1.20) and an unadjusted risk difference (RD) of 10.0 (95% CI, 3.5-16.5) per 10 000 births. Among exposed pregnancies, the risks of oral clefts were 14.0 (95% CI, 11.6-16.5) per 10 000 vs 11.1 (95% CI, 10.6-11.6) per 10 000 unexposed pregnancies (estimated based on 124 exposed and 1912 unexposed cases), resulting in an RR of 1.26 (95% CI, 1.05-1.51) and an RD of 2.9 (95% CI, 0.4-5.4) per 10 000 births (Table 2 and Figure 1).
The increase in risk of cardiac malformations associated with ondansetron in the unadjusted analyses was related to ventricular septal defects (400 exposed and 6826 unexposed cases; RR, 1.14; 95% CI, 1.04 to 1.27; RD, 5.7; 1.2 to 10.2 per 10 000) and secundum atrial septal defects (216 exposed and 3080 unexposed cases; RR, 1.37; 95% CI, 1.19 to 1.57; RD, 6.6; 95% CI, 3.3 to 9.9 per 10 000). The other specific cardiac malformations were less common, ranging between less than 11 and 113 exposed cases, and none of the unadjusted RR suggested an increased risk. The increased risk for oral clefts was attributable to cleft palate (65 exposed; 988 unexposed cases; RR, 1.29; 95% CI, 1.00 to 1.65; RD, 1.6; −0.2 to 3.5 per 10 000). There was no evidence of an increased risk for cleft lip (33 exposed; 620 unexposed cases; RR, 1.04; 95% CI, 0.73 to 1.48; RD, 0.1; 95% CI, −1.2 to 1.4 per 10 000) or cleft lip with cleft palate (48 exposed vs 925 unexposed cases; RR, 1.01; 0.76 to 1.35; RD, 0.1, 95% CI, −1.5 to 1.6 per 10 000) in unadjusted analyses.
The risk for any congenital malformation was 370.4 (95% CI, 358.0 to 382.9) per 10 000 ondansetron-exposed pregnancies compared with 313.5 (95% CI, 310.9 to 316.1) per 10 000 unexposed pregnancies (estimated based on 3277 exposed and 54 174 unexposed cases), corresponding to an RR of 1.18 (95% CI, 1.14 to 1.22) and an RD of 56.9 (95% CI, 44.2 to 69.6) per 10 000 births (eTable 4-5 in the Supplement).
After stratification on the propensity score, all measured patient characteristics were balanced between the ondansetron-exposed and unexposed groups judged by an absolute standardized difference of less than 0.1 (Table 1; eTable 3 in the Supplement). Although adjusting for the treatment indications and associated factors (propensity score level 1) did not substantially change the crude risk estimates, accounting for all prespecified potential confounding variables (propensity score level 2) resulted in a null point estimate for cardiac malformations (RR, 0.99; 95% CI, 0.93 to 1.06; RD, −0.8; 95% CI, −7.3 to 5.7 per 10 000 births) and for the secondary outcome of congenital malformations overall (RR, 1.01; 95% CI, 0.98 to 1.05; RD, 5.4; 95% CI, −7.3 to 18.2 per 10 000 births), and a statistically significant RR of 1.24 (95% CI, 1.03 to 1.48) for oral clefts corresponding to an RD of 2.7 (95% CI, 0.2 to 5.2) per 10 000 births. These findings were confirmed in analyses that adjusted for proxies for unmeasured confounders through use of high-dimensional propensity scores (Figure 1).
A total of 185 876 women filled a prescription for antiemetics other than ondansetron during the first trimester: 52 818 for metoclopramide, 139 932 for promethazine, and 9036 for pyridoxine (eTables 6-9 in the Supplement). Results when using women exposed to other antiemetics as the reference group were consistent with the main analyses (adjusted RR for cardiac malformations, 1.01; 95% CI, 0.92-1.12; for oral clefts, adjusted RR, 1.32; 95% CI, 1.03-1.70; and for any congenital malformation RR, 1.00; 95% CI, 0.95-1.05; Figure 2). Results were similar across the individual antiemetic agents (Figure 2 and eTables 10-13 in the Supplement).
The findings generally were not sensitive to changes in the exposure definition (≥2 prescriptions, 6-12 weeks). Requiring 2 or more prescriptions to be filled during the first trimester did not strengthen the association for oral clefts (adjusted RR, 1.15; 95% CI, 0.85-1.56). In the negative control analyses, which evaluated the association with exposure to ondansetron after the etiologically relevant time window, no increased risk was observed for oral clefts (RR, 0.98; 95% CI, 0.66-1.45). When using these women with exposure to ondansetron after the etiologically relevant window as an alternative reference group, adjusted estimates were consistent with the main analyses (Figure 2 and eTables 14-17 in the Supplement).
We quantified the potential for selection bias due to the restriction of the cohort to livebirths. The most extreme scenario considered was a probability of livebirth of 80% among unexposed infants without a malformation, 55% among unexposed infants with a cardiac or any type of malformation, 70% among unexposed infants with oral clefts, and a 20% absolute decrease in the probability of livebirth in those exposed to ondansetron compared with those unexposed for both malformed and nonmalformed fetuses. Based on these assumptions, the RR estimate would shift from 0.99 to 1.17 for cardiac malformations, 1.24 to 1.30 for oral clefts, and 1.01 to 1.19 for any malformation (see the eAppendix in the Supplement for details in the Supplement).
In exploratory analyses, we screened for associations with other specific malformations (eTable 4 in the Supplement). In the context of multiple comparisons, the risk of ear malformations was elevated: the absolute risks were 3.8 (95% CI, 2.7-5.4) per 10 000 exposed and 2.4 (95% CI, 2.2-2.7) per 10 000 unexposed pregnancies, corresponding to an adjusted RR of 1.64 (95% CI, 1.16-2.33) and an RD of 1.5 (95% CI, 0.2-2.8) per 10 000 births.
Quiz Ref IDIn this national cohort of publicly insured pregnant women, there was no association between exposure to ondansetron during the first trimester of pregnancy and increased risk of cardiac malformations or congenital malformations overall, after accounting for potentially confounding conditions. Given the large cohort size, the estimates were precise with an upper bound of the 95% CI of the adjusted relative risk of 1.06 for cardiac malformations and of 1.05 for congenital malformations overall. There was a small increase in the risk of oral clefts associated with ondansetron exposure with the upper bound of the 95% CI of the adjusted relative risk at 1.48, corresponding to 5 additional cases per 10 000 prenatally exposed livebirths. These findings were consistent across a broad range of sensitivity analyses.
This study expands on the evidence available to date. The largest and most recent cohort study published in the peer-reviewed literature included 19 cardiac malformations, 1 oral cleft, and 49 malformations overall among 1349 exposed women (eTable 1 in the Supplement).31 Conclusions about the doubling of risk of cleft palate in a case-control study conducted by the Centers for Disease Control & Prevention using data from 1997 through 2004 were based on 11 exposed cases (odds ratio, 2.4; 95% CI, 1.2-4.8).10 The risk was attenuated but remained increased in more recent analyses using the same data source for the years 2005 through 2014 based on 40 exposed cases (odds ratio, 1.6; 95% CI, 1.1-2.4).12 However, in another case-control study using data from the Slone Epidemiology Center Birth Defects Study for 1997 through 2014 no increase in risk was observed based on 11 exposed cases (odds ratio, 0.5; 95% CI. 0.3-1.0).12
Quiz Ref IDIn some prior studies suggesting an increase in cardiac malformations or cleft palate, exposure assessment was completely10,12 or partially31 based on women’s self-report, which may have resulted in bias since women who gave birth to an infant with a congenital malformation may be more likely to report all their exposures than women who gave birth to a healthy child (eTable 1 in the Supplement). Other studies did not account for important potential confounding variables, such as the underlying indication or its severity, maternal comorbidities (eg, diabetes) and concomitant medication use.8,9,31 In addition, some of the previously reported associations have been found in the context of multiple comparisons,10 increasing the likelihood of a chance finding. For this same reason, a cautious interpretation of the finding of a potential increase in the risk of ear deformities that emerged in exploratory analyses is warranted. It should be viewed as a potential safety signal that requires replication in other studies. While awaiting additional data, it is important to recognize that ear deformities are uncommon, with a base prevalence of 2 per 10 000 unexposed pregnancies in our cohort.
Apart from the large cohort size, this study has several strengths. First, information on medication exposures was collected in a prospective manner based on filled prescriptions for medications and is thus free from recall bias. Second, the MAX data contain rich patient-level information for confounding control, including maternal demographic characteristics, medical conditions, and medication exposures. Third, in addition to comparing exposed to unexposed women, ondansetron-exposed women were compared with women exposed to other antiemetics and consistent results were documented. Aside from reducing the likelihood of confounding, this analysis also addressed the clinically relevant question of which treatment is preferred for women whose symptoms are not self-limiting and do not resolve with nonpharmacological conservative measures. Fourth, both cardiac malformations and oral clefts have been shown to have a high positive predictive value when identified using computerized records. The positive predictive value for cardiac malformations was previously estimated at 78%,20 and some well-known associations were replicated in the MAX data set using this definition (ie, maternal diabetes, anticonvulsant use, multifetal pregnancy)15 providing confidence in the specificity of the outcome definition. The positive predictive value for oral clefts has been estimated to be between 93% and 97%.21,32 Such modest nondifferential misclassification would minimally bias the RR toward the null.
This study has several limitations. First, filling a prescription does not mean that the medication was actually consumed, which could bias the results toward the null. To address this limitation, sensitivity analyses in which women were required to have filled at least 2 prescriptions during the first trimester were conducted based on the notion that if a woman refills a prescription she is likely to have consumed the prescribed medication. This approach did not result in stronger associations. Second, in nonrandomized studies, there is always concern about residual confounding due to unmeasured or poorly measured characteristics. For example, absence of a recorded diagnosis is equated with absence of the disease. It is possible, however, that the clinician did not record the diagnosis. This may have resulted in some misclassification of the confounder information and hence affect our ability to control for confounding. Although negative residual confounding is typically not a concern for null findings because drug exposure is not expected to be associated with factors protective against congenital malformations (ie, cardiac malformations or congenital malformations overall), positive residual confounding could be a potential explanation for the increased risk in oral clefts. An attempt was made to refute this alternative explanation using different strategies including adjustment for proxies of unmeasured confounders through high-dimensional propensity scores, use of alternate reference groups, and a negative control analysis. No increased risk of oral clefts was observed using the negative control exposure window, supporting the validity of this association. Third, because the cohort was restricted to livebirths, severe congenital malformations that result in pregnancy losses or terminations will be missed. Nonsyndromic oral clefts do not result in fetal deaths and are rarely a reason for terminations.33 It is therefore highly unlikely that a differential proportion of nonlive births among ondansetron users vs the reference groups would have biased our findings. For cardiac malformations and any congenital malformation, formal quantification of the potential for selection bias revealed that the differences in the proportion of nonlive births among ondansetron users vs the reference groups within levels of covariates used in the adjustment would have to be unrealistically strong in order to dilute a clinically meaningful increase in risk.
Quiz Ref IDFourth, this study was focused on individuals with Medicaid insurance. Medicaid covers the medical expenses for approximately 50% of all pregnancies in the United States,34 making this an important population to study. The cohort inclusion criteria resulted in the selection of a more disadvantaged subpopulation within Medicaid, mostly composed of low-income adults, multiparae, and women with disabilities. The characteristics of this Medicaid-eligible population of pregnant women, ie, a young, racially diverse, and vulnerable population with a high burden of disabilities, are not expected to affect the biological relations studied. Therefore, the results should be generalizable to the broader Medicaid population, as well as to commercially insured pregnant women in the United States, and pregnant populations in other countries. Fifth, the study was based on data from 2000 through 2013, but the biological association between ondansetron exposure and malformations should not change over time.
Among offspring of mothers enrolled in Medicaid, first-trimester exposure to ondansetron was not associated with cardiac malformations or congenital malformations overall after accounting for measured confounders but was associated with a small increased risk of oral clefts.
Corresponding Author: Krista F. Huybrechts, MS, PhD, Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital, 1620 Tremont St, Ste 3030, Boston, MA 02120 (firstname.lastname@example.org).
Accepted for Publication: October 26, 2018.
Author Contributions: Dr Huybrechts had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Huybrechts, Hernández-Díaz, Gray, Patorno, Desai, Bateman.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Huybrechts, Hernández-Díaz, Straub, Bateman.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Huybrechts, Hernández-Díaz, Straub, Mogun, Bateman.
Obtained funding: Huybrechts, Hernández-Díaz, Bateman.
Administrative, technical, or material support: Huybrechts, Straub, Bateman.
Supervision: Huybrechts, Hernández-Díaz, Bateman.
Conflict of Interest Disclosures: Dr Huybrechts reports receiving research grants to her institution from Eli Lilly, Pfizer, GlaxoSmithKline, and Boehringer-Ingelheim. Dr Patorno reports receiving research grants to her institution from Boehringer Ingelheim and GlaxoSmithKline. Dr Desai reports receiving grants to his institution from Merck, Bayer, and Vertex. Dr Bateman reports receiving grants to his institution from Eli Lilly, GlaxoSmithKline, Pacira, Baxalta, Pfizer, and Aetion and having served on an expert panel for a postpartum hemorrhage quality improvement project that was conducted by the Association of Women's Health, Obstetric, and Neonatal Nurses and funded by a grant from Merck for Mothers. Dr Hernández-Díaz reports receiving research grants to her institution from Eli Lilly, GlaxoSmithKline, and the National Institutes of Health and consulting fees from Roche and having served as an epidemiologist with the North America AED pregnancy registry, which is funded by multiple companies. Dr Gray reports consulting for Quest Diagnostics. No other conflicts of interest were reported.
Funding/Support: This study was supported by grant R03 HD091699 from the National Institute of Child Health and Human Development. Dr Huybrechts was supported by career development grant K01MH099141 from the National Institute of Mental Health. Dr Bateman was supported by career development grant K08HD075831 from the National Institute of Child Health and Human Development. Dr Gray was supported by career and development grant NIH K12 BIRCWH from the National Institutes of Health to Harvard Medical School. Dr Patorno was supported by career development grant K08AG055670 from the National Institute on Aging.
Role of the Funder/Sponsor: The National Institutes of Child Health and Human Development had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: