First trimester ondansetron exposure and risk of structural birth defects

Accepted Manuscript Title: First Trimester Ondansetron Exposure and Risk of Structural Birth Defects Authors: April Zambelli-Weiner, Christina Via, Matt Yuen, Daniel J. Weiner, Russell S. Kirby PII: DOI: Reference:

S0890-6238(18)30123-0 https://doi.org/10.1016/j.reprotox.2018.10.010 RTX 7753

To appear in:

Reproductive Toxicology

Received date: Revised date: Accepted date:

30-3-2018 13-8-2018 26-10-2018

Please cite this article as: Zambelli-Weiner A, Via C, Yuen M, Weiner DJ, Kirby RS, First Trimester Ondansetron Exposure and Risk of Structural Birth Defects, Reproductive Toxicology (2018), https://doi.org/10.1016/j.reprotox.2018.10.010 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

FIRST TRIMESTER ONDANSETRON EXPOSURE AND RISK OF STRUCTURAL BIRTH DEFECTS April Zambelli-Weiner1, Christina Via1, Matt Yuen1, Daniel J. Weiner1, Russell S. Kirby2 1

SC RI PT

TTi Health Research & Economics 4500 Black Rock Road Cedarbrook Center, Suite 201 Hampstead, MD, USA 21074 2

Department of Community and Family Health, College of Public Health, University of South Florida 13201 Bruce B Downs Blvd Tampa, FL 33612

M

A

N

U

Corresponding Author: April Zambelli-Weiner, PhD TTi Health Research & Economics 4500 Black Rock Road Cedarbrook Center, Suite 201 Hampstead, MD, USA 21074 Telephone: (800) 580-2990 Email: [email protected]

EP

TE

Our study examined ondansetron exposure and risk of structural birth defects. This large, U.S.-based study addresses limitations of prior epidemiological studies. Statistically significant increased risk was shown for specific structural birth defects. Methods address misclassification bias often present in administrative claims data. Results support evidence of teratogenicity of ondansetron.

A

CC

    

D

Highlights for “First trimester ondansetron exposure and risk of structural birth defects” article.

Word count: 150 ABSTRACT: This study investigates risk of specific structural birth defects associated with ondansetron exposure during the first trimester in a large US commercially-insured population.

SC RI PT

Medical claims data were obtained from Truven Health Analytics for 864,083 mother-infant pairs from 2000 to 2014. Logistic regression was used to measure the association between first trimester exposure to ondansetron and risk of cardiac defects, orofacial clefts and other specific structural defects in offspring. First trimester exposure to ondansetron was associated with increased risk of cardiac (OR: 1.52 95% CI: 1.35-1.70) and orofacial cleft defects (OR: 1.32 95% CI: 0.76-2.28) in offspring compared to women with no antiemetic exposure during pregnancy.

A

N

U

This analysis addresses limitations of prior studies including limited power, exposure misclassification, and generalizability to the US population. In a large, US population we found a statistically significant association between early pregnancy ondansetron exposure and specific structural birth defects in offspring.

M

KEYWORDS: ondansetron; Zofran; birth defects; congenital malformations; congenital anomalies; cardiac defects; orofacial clefts; nausea and vomiting in pregnancy

D

INTRODUCTION

TE

1

Nausea and vomiting in pregnancy (NVP) is a frequent complaint during first trimester. In recent

EP

studies 70-90% [1,2] of women report experiencing NVP, with 0.5-1.5% [3,4] diagnosed with the more severe hyperemesis gravidarum (HG). Approaches to managing NVP and HG include

CC

natural remedies such as ginger [5–7], over-the-counter drugs including antihistamines [6,8,9], and prescription pharmaceutical treatments [6,10], with the use of prescription medications for NVP

A

increasing in recent years [11,12]. The 2004 clinical guidelines from the American College of Gynecology (ACOG), which were in use during the time of our study, recommend five drugs for the treatment of NVP: ondansetron, Diclegis® (doxylamine succinate and pyridoxine Hydrochloride), metoclopramide, promethazine, and methylprednisolone. In the most recent 2018

2

ACOG guidelines, the same five drugs remain with several additions (prochloroperazine, chlorpromazine, and trimethobenzamide) [10,13]. While ondansetron maintains its position in the clinical guidance documents as a third-line pharmacologic therapy, Taylor and colleagues found

SC RI PT

that, as of 2014, ondansetron represents the most frequently prescribed medication for the treatment of NVP and HG in the United States with use increasing from <1% of pregnancies in 2001 to over 22% in 2014 [12].

Prior studies on the fetal safety of ondansetron have produced varied results, reflecting

U

heterogeneity in study populations, methodological limitations, and small sample sizes. Studies

N

such as those done by Einarson et al 2004, Asker et al 2005, Colvin et al 2013, Pasternak et al

A

2013, and Parker et al 2018 found no increased risk of major birth defects, while Danielsson et al

M

2014 and Anderka et al 2012 found increased risks of cardiac defects and cleft palates, respectively [14–19, 21]. Translation of the evidence base to fuel effective policy and clinical decision making

D

has been hampered by the heterogeneity and limitations in currently published epidemiological

TE

studies. For example, some studies have been limited by small numbers of exposed patients,

EP

lumping of birth defects, and misclassification bias due to patient recall of exposure. In addition, many have been conducted in non-US populations, where clinical care and prescribing patterns

CC

can differ significantly from the U.S., limiting generalizability. The two major limitations that have hampered many studies are small sample sizes (i.e. inadequate power, particularly for

A

important subgroup analyses) and risk of bias from exposure misclassification due to reliance on filled prescriptions as a surrogate of exposure. It was our goal to directly address these limitations and shed new light on the evidence base by (1) amassing a large enough sample to be sufficiently powered to address the limitations of prior studies and (2) directly address and quantitatively

3

estimate the magnitude of exposure misclassification bias introduced by the reliance on ondansetron filled prescription data.

SC RI PT

To address these limitations and gaps in the literature, we conducted a nested case control study using a large US administrative claims database spanning 15 years to study the association between ondansetron use during first trimester and risk of specific structural congenital malformations, while considering the impact of important potential biases and modifying factors such as

MATERIALS AND METHODS

N

2

U

confounding by indication, exposure misclassification and concomitant drug exposure.

A

2.1 2.1 DATABASE

M

We conducted a nested case-control study using a large, proprietary, US administrative health care

D

database, the Truven Health MarketScan Commercial Database [26]. This database captures the

TE

full episode of reimbursable care for each patient during plan enrollment, including inpatient and outpatient medical care, prescription drug use and other resource utilization. The MarketScan

EP

Commercial database is comprised of adjudicated data mostly from self-insured companies. From each of these companies, an enrollment file was received, where all dependents are linked to the

CC

employee. The enrollment file is used to link the mother/infant pairings. For the study, pregnant

A

mothers are identified by their delivery. Using that as an anchor, the different claims based on the mother and infant's IDs were tracked. Diagnosis and procedure codes are identified by the International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM) and Current Procedural Terminology (CPT) codes.

Drug exposures were identified through the

specific compilation of lists derived from The National Drug Code Directory. Institutional Review 4

Board approval was not required, as this study used existing, fully de-identified data and as such is exempt from 45 CFR 46 requirements.

SC RI PT

2.2 STUDY POPULATION The source population included mother-child pairs resulting from all live births from 2000 to 2014 who had one year of follow-up for the infant(s). Mothers were eligible if they were continuously enrolled for 16 months prior to delivery and were between 15 and 49 years old on the date of delivery. Mother-child pairs were excluded if there was an increased baseline risk of congenital

U

malformations due to previous offspring with chromosomal birth defects or exposure to known

N

teratogens, defined as: a recently recorded maternal diagnosis in the 16-month pre-birth period of

A

chromosomal anomalies (ICD-9 758.xx), toxoplasmosis, other (syphilis, varicella-zoster,

M

parovirus B19), rubella, cytomegalovirus, and herpes (TORCH) infections, or if the mother filled

D

a prescription for thalidomide or isotretinoin in the pre-birth period. Other exclusion criteria

TE

included exposure to an antiemetic other than ondansetron, including Diclegis®, metoclopramide, promethazine, or methylprednisolone anytime during pregnancy.

Those with ondansetron

EP

exposure exclusively outside the first trimester were also excluded to minimize exposure

CC

misclassification.

A

2.3 EXPOSURE MEASUREMENT Exposure to ondansetron was defined as a filled prescription for ondansetron or medical administration of ondansetron in the medical office or hospital setting during first trimester and was identified using NDC codes and trade names. The period of first trimester exposure was defined as the period from the beginning of the estimated date of conception to the estimated end

5

of the first trimester (91 days following the estimated conception date), specifically 287 to 147 days prior to delivery, based on an estimated conception period of 287-252 days prior to delivery for a singleton birth and 273-238 days prior to delivery for a multiple birth based on a previously

SC RI PT

published algorithm using the Truven Marketscan data [27].

Mother-child pairs were considered exposed if they had any first trimester exposure, which is a filled prescription for ondansetron or a claim for medical administration of ondansetron. To control for concomitant antiemetic use, ondansetron exposed mother-child pairs were defined as

U

those exclusively exposed to ondansetron during first trimester, meaning there was not a claim for

N

Diclegis®, metoclopramide, promethazine or methylprednisolone during the first trimester period.

A

Because antiemetics are prescribed prophylactically to be used on an “as needed” basis, there is

M

significant risk of exposure misclassification in prescription data. Due to this, our primary analysis examined medical administration of ondansetron, where the risk of exposure misclassification is Secondary analyses were conducted using all ondansetron exposure

D

very low if not null.

TE

(prescription claims plus medical administration) to quantitatively evaluate the impact of

EP

misclassification bias on the association when prescription claims data are utilized. Mother-child pairs were considered unexposed if they had no exposure to any antiemetics listed in the 2004

CC

ACOG guidelines anytime during the pregnancy.

A

2.4 BIRTH OUTCOMES Outcomes were identified using ICD-9-CM codes (see supplemental appendix). Cases were identified as having one or more claims with a relevant diagnosis code within 365 days of the date of birth. Birth defects were chosen a priori based on previous studies, animal models, and expert

6

opinions. Primary outcomes of interest were cardiac defects (ICD-9 745.xx) and orofacial cleft defects (ICD-9 749.xx). Secondary outcomes include specific cardiac and orofacial cleft defects, as well as other types of congenital heart anomalies; hypoplastic left heart syndrome; congenital

SC RI PT

anomalies of the circulatory system; anomalies of the larynx, trachea, and bronchus; anencephalus; spina bifida without anencephaly; limb reduction defects; craniosynostosis; congenital diaphragmatic hernia; and renal collecting system anomalies. A negative control group was identified, containing all birth defects with ICD-9 codes 740.xx-759.xx, excluding the a priori birth defects of interest.

U

2.5 COVARIATES

N

Covariates identified and available in our database included: maternal age at birth, infant year of

A

birth, infant gender, US region of birth, maternal medical history (obesity, diabetes mellitus,

M

epilepsy, hypertension, cancer), and medications taken in first trimester including initiation of

D

acid-reducing therapies, psychotropics, prescription folic acid, macrolide antibiotics,

TE

corticosteroids, and anticonvulsants. Low birth weight, multiple gestation, high risk pregnancy diagnosis, prior preterm delivery, family history of birth defects, and diagnosis of NVP or HG

EP

were also collected, but were difficult to assess from ICD-9 codes alone, due to underutilization.

CC

True confounders, by definition, must be associated with both the exposure and the outcome under investigation, but cannot be in the causal pathway or an intermediate (i.e., influenced by exposure).

A

Potential confounders were first identified through a review of the literature and the team’s understanding of the clinical care of pregnant women and the epidemiology of birth defects. Confounding was then evaluated by adding the variable to the base exposure-outcome model and evaluating the change in the risk estimate. If the risk estimate changed by ≥10% this indicated the presence of confounding and that variable was included in the multivariate models [28,29]. 7

Maternal medical history, comorbidities and other medications were measured from medical and prescription claims during the pre-birth period. Comorbid conditions were flagged as present if a

SC RI PT

patient had ≥ 1 medical claim with an ICD-9-CM diagnosis code for the condition. Additional medication use was identified if a patient had ≥ 1 HCPCS claim or HCPCS code associated with the medication in the first trimester period, as previously described.

2.6 STATISTICAL ANALYSIS

U

The statistical analyses were performed using Stata/MP 13.1 (College Station, TX; StataCorp LP).

N

Characteristics of mother-child pairs were compared by birth outcome using chi-square test or

A

Fisher’s exact test for categorical variables and Student’s t-test for continuous variables. Logistic

M

regression models were used to test for a statistical association between first trimester ondansetron use and risk of cardiac defects and orofacial cleft defects, as well as specific cardiac, orofacial and

TE

D

other structural birth defects, and finally a negative control of “other birth defects.” Prevalence

EP

odds ratios and 95% confidence intervals were calculated.

Sensitivity analyses were conducted to evaluate the impact of potential confounding by indication

CC

on the findings and to evaluate potential ascertainment bias and the potential for uncontrolled confounding bias. Specifically, confounding by indication was addressed by comparing the

A

exposed cohort to only those with both no antiemetic exposure during pregnancy and a diagnosis of NVP/HG. Ascertainment bias occurs when the probability of detecting a case is systematically different between comparison groups. In this case, publicity around a drug’s safety can lead to enhanced surveillance, or screening, for birth defects in women who took the drug. To evaluate

8

for potential ascertainment bias we conducted a stratified analysis by year, with yearly cut-off points relating to increased awareness/publicity of possible risks associated with taking

3

SC RI PT

ondansetron in the first trimester.

RESULTS

3.1

STUDY POPULATION

After exclusions, a total of 864,083 mother-child pairs were identified. Early exposure to ondansetron occurred in 76,330 mother-child pairs (8.8%), and early exposure to medical

N

U

administration of ondansetron occurred in 5,557 mother-child pairs (0.64%).

A

Table 1 shows the characteristics of women by birth outcome of the child. There were 802,253

M

infants with no birth defects, 32,100 infants were diagnosed with cardiovascular birth defects, and

D

1,590 infants were diagnosed with orofacial cleft defects. Because of the large sample sizes, even

TE

small differences in covariate distributions between birth defect cohorts were statistically significant. However, when potential confounders were evaluated by adding the variable to the

EP

base exposure-outcome model, none of them reached the a priori 10% change threshold. However, to facilitate comparison of results across studies we also present results adjusted for

CC

infant year of birth, infant gender, and mother’s age at infant birth for each model in the main and

A

secondary analyses.

3.2 PRIMARY ANALYSIS Table 2 shows the association of first trimester exposure to ondansetron with any cardiac defects, and any orofacial cleft defects, both with and without adjustment for mother’s age at birth, infant 9

gender, and infant year of birth. The primary results, in which bias from exposure misclassification is minimized (medical administration of ondansetron), demonstrate that first trimester ondansetron exposure is associated with statistically significantly increased risk for all cardiac defects (OR:

SC RI PT

1.52, 95% CI: 1.35-1.70). First trimester ondansetron exposure was associated with an increased risk of orofacial cleft defects (OR: 1.32, 95% CI: 0.76-2.28), although there is significant variability around the estimates due to small sample sizes.

3.3

SECONDARY ANALYSES

U

Table 2 also shows association of first trimester ondansetron exposure and specific structural birth

N

defects. Ondansetron exposure was associated with a significantly increased risk of all specific

A

cardiac birth defects, and risk estimates were especially elevated for atrial septal defects (OR: 1.62,

M

95% CI: 1.43-1.84) and atrioventricular septal defects (OR: 2.68, 95% CI: 1.61-4.47). Ondansetron

D

exposure during first trimester was also associated with significantly elevated risk of other

TE

circulatory defects (OR: 1.83, 95% CI: 1.45-2.29), diaphragmatic hernia (OR: 2.49, 95% CI: 1.185.25), and renal collecting system anomalies (OR: 1.28, 95% CI: 1.01-1.63).

EP

3.4 SENSITIVITY ANALYSES

CC

Table 3 shows results from the two sensitivity analyses. Restricting the control (unexposed) group to only those mother-child pairs with a diagnosis of NVP or HG did not change the results

A

appreciably (cardiac defects: OR: 1.51, 95% CI: 1.24-1.85 versus OR: 1.52, 95% CI: 1.35-1.70). Similarly, stratifying the results by year does not show significantly different risk estimates before and after key information was released on the safety of ondansetron. Infants conceived between 2000 to 2006, 2007 to 2011, and 2012 to 2014 all had elevated risks of cardiac defects if exposed

10

to early ondansetron exposure with odds ratios of 1.66 (95% CI: 0.8-3.24), 1.29 (95% CI: 1.091.52), and 1.56 (95% CI: 1.32-1.85) respectively.

DISCUSSION

SC RI PT

4

Our nested case control study demonstrates a statistically significant association between first trimester exposure to ondansetron and a range of specific structural birth defects in offspring in a large, US population when exposure misclassification is minimized. Further, our data show a high proportion of women receiving ondansetron for control of NVP and HG in pregnancy (20% in

U

2014), comparable to Taylor et al 2017 [12], as well as a significant number of women receiving

N

only ondansetron, indicating that ondansetron is being used as first-line therapy contrary to clinical

A

guidelines. Previous studies have demonstrated an increased risk of cardiovascular defects [15,22]

M

and septal defects [15,22], with particularly elevated risks for specific septal defects [22]. Data

D

has been more limited in the US with studies showing increased risk of cleft palate [16] and

TE

clubfoot [23].

EP

We sought to address concerns around limitations present in previous studies that may be confounding translation of the evidence base on the fetal safety of ondansetron, including: recall

CC

bias, exposure misclassification, low exposure prevalence, and small sample sizes that reduce study power and may have precluded important subgroup analyses. One of the largest risks of

A

bias to pharmacoepidemiological studies utilizing prescription claims as a marker of exposure is that the prescription may not be representative of actual exposure. We addressed this risk of bias by setting a more stringent exposure definition where the probability of exposure is near certain: ondansetron administered in a medical or hospital setting. By presenting results for both combined

11

medical and prescription claims and medical claims alone, the impact of misclassification of exposure becomes clear, revealing a significant biasing of the risk estimate towards the null using prescription data – and illustrating the point previously made by Danielsson and colleagues that

SC RI PT

“as in all such studies, including ours, it is not known if the women who filled a prescription actually took the drug during the organogenetic period, a phenomenon which results in a bias of the risk estimate towards null” [15].

Another potential risk of bias in pharmacoepidemiological studies is confounding by indication,

U

in which the indication for the drug itself may be causally related to the adverse event, and

N

therefore the pharmacologic treatment appears harmful because it’s only given to sick people. We

A

considered the possibility that medical administration of ondansetron was a surrogate for a more

M

severe case of NVP or HG, thus raising concerns about confounding by indication. However, research has shown that NVP may favorably affect neonatal outcomes. A systematic

D

review showed that mothers who experience NVP during their pregnancy had a decreased risk of

TE

congenital abnormalities in the newborn [30]. Additionally, NVP has been linked to decreased

EP

risk of miscarriages, premature birth, and small for gestational age newborns, suggesting that NVP may have a protective effect on the fetus [30–32]. The implication is that in all epidemiological

CC

studies the true risk of ondansetron may be underestimated if there is a protective effect of the indication and it is not controlled for. In addition, a recent study specifically addressing medical

A

administration of ondansetron reported that among women who received medical administration of ondansetron as a first line therapy, the majority did not have a history of NVP in pregnancy (91.4%) and were not currently being treated for NVP (60.3%) [33]. To address possible confounding by indication, we conducted a sensitivity analysis in which those exposed to

12

ondansetron were compared to a subgroup of unexposed mother-child pairs with a specific diagnosis of NVP or HG. Controlling for indication increased risk estimates across the board, most significantly so in adjusted analyses, supporting the notion that NVP/HG may exert

SC RI PT

protective effects.

Another potential source of bias is ascertainment bias. Our study cohort spans 15 years, and during this time it is possible that certain events may have led to increased awareness/publicity of possible risks associated with taking ondansetron during first trimester. In a scenario of ascertainment bias,

U

awareness and anxiety associated with ondansetron may increase, causing women who have taken

N

ondansetron to be more likely to undergo diagnostic tests that detect malformations. We addressed

A

this potential risk of bias by stratifying our analyses to correspond with dates related to increasing

M

publicity around the safety of ondansetron: (1) 2006, representing the first study published showing ondansetron crosses the placenta [34]; and (2) 2012, representing the FDA warning

D

regarding risks of adverse events when using ondansetron in late 2011 and a CDC funded study

TE

finding association between ondansetron risk and cleft palate in early 2012 [16,35]. Finally, to

EP

overcome potential detection bias, infants in our study were followed for one year post-delivery to

CC

allow latent diagnosis of birth defects.

As with all studies, ours is not without limitations. While studies employing administrative claims

A

databases can be valuable due to the high sample size achievable, they have inherent limitations. The inability to control for all potential confounders is a limitation of epidemiological research and of administrative claims databases, in particular.

Data were unavailable on maternal

sociodemographic variables such as race, education and parity as well as lifestyle risk factors such

13

as smoking, alcohol consumption and over the counter drug use in the pre-conception and first trimester periods. To address this, sensitivity analyses using the method of Greenland [36] were conducted to evaluate the impact of uncontrolled confounding on risk estimates. Taking smoking

SC RI PT

as an example, the strength of the relationship between smoking and the probability of both the exposure and outcome is modest and adjustment for smoking does not meaningfully alter the risk estimates (data provided in supplemental appendix). Given that many of the factors that could not be reliably adjusted for either due to lack of data or missing data (maternal comorbidities and socioeconomic factors) are likely not strong confounders (strongly associated with both the

U

probability of ondansetron exposure and the risk of birth defects), we do not believe that bias from

N

uncontrolled confounding is of significant concern. The results from our negative control indicate

A

that there may be some modest residual confounding that is captured by the adjusted model, but

M

adjustment does not significantly alter the magnitude and direction of the risk estimates.

D

Another limitation intrinsic to administrative claims data is that the conception date must be

TE

estimated, which can lead to misclassification of the relevant exposure window (first trimester).

EP

We attempted to minimize this type of exposure misclassification by excluding mother-child pairs who were only exposed to ondanestron after the first trimester, or 140 days or less before birth.

CC

While there remains a small probability of exposure misclassification (that exposure occurred outside the key developmental window), the implication is that mother-child pairs are being

A

labeled as exposed when they are not exposed, leading risk estimates to be biased toward the null.

Additional limitations include the inability to include terminations, conduct dose-response analyses and potential misclassification of specific birth defect outcomes, each discussed briefly

14

here. Because the study cohort is based on live births, pregnancies that were terminated due to fetal anomalies could not be included, and those infants who didn’t survive their birth defect up to one year of age or were switched to an insurance plan not included in the Truven MarketScan

SC RI PT

database could not be included. Dose response analysis was prohibitive given that antiemetics can be taken on an “as needed” basis and because days-supply associated with prescriptions is not always uniformly recorded in administrative claims data [25]. Despite unreliability of dose data within the database, there is no data supporting the notion that medical IV administration represents a higher dose than oral administration. According to UptoDate, a common peer-

U

reviewed resource for hospital physicians, the dose of ondansetron is the same regardless of route

N

[37]. Finally, we had a higher rate of cardiac birth defects than expected. A recent study by

A

Huybrechts et al 2018 found 1.3% of their control group had cardiac birth defects, compared to

M

3.7% of our control population [38]. Diagnostic codes in administrative health databases can be problematic for particular outcomes. Although the use of these codes generally indicates that a

D

clinical diagnosis has been made to support the code, with birth defects there are situations where

TE

these codes may be used for other purposes. For example, a generic code for congenital heart

EP

defects (for example ASD or VSD) may be noted as a justification for ordering an electrocardiogram to rule out a heart lesion. These codes are frequently not removed from the

CC

record. A 2005 study further confirms this, demonstrating that reliance on ICD-9 cardiac defect codes can lead to false positives [39]. However, it is unlikely that this would happen more often

A

in the ondansetron exposed group than the unexposed group, and therefore any resultant bias due to outcome misclassification would only serve to underestimate the true relative risk.

15

Our results contrast the current prevailing view that there is only a low, if any, increased risk between first trimester ondansetron exposure and risk of birth defects in offspring [13,40]. Our results in a US population do show consistency with some results from previous studies examining

SC RI PT

the relationship between first trimester ondansetron exposure and risk of birth defects in offspring. Specifically, Danielsson and colleagues [15] report elevated risks for cardiovascular defects (OR: 1.62, 95% CI: 1.04-2.14) and septal defects (OR: 2.05, 95% CI: 1.19-3.28)[15] comparable to those reported in our study (OR: 1.52, 95% CI: 1.35-1.701.35-1.70; OR: 1.53, 95% CI: 1.36-1.71, respectively). Unadjusted data from Pasternak and colleagues show cardiac birth defect risk to be

U

highest for atrioventricular septal defects (OR: 4.8, 95% CI: 0.25-63.91), which are similarly

N

highest in our study (OR: 2.68, 95% CI: 1.61-4.47) [19]. Similar trends are observed in an abstract

M

A

by Andersen and colleagues [22].

In contrast, other epidemiological studies failed to detect a statistically significant association

D

between ondansetron and specific birth defects [19–21,41,42] most likely due to limitations in

TE

study design or analysis. For example, some studies lumped all birth defects into a binary

EP

categorical variable without considering specific birth defects [19,21]. Lumping birth defects can mask the effects of drug exposures leading to misclassification bias toward the null [43], as

CC

demonstrated by studies that found no association in lumped categories but positive association with specific defects in the same study [19,21,41]. Many of the studies employed large

A

administrative claims data or prescriptions registers, such as Pasternak 2013, relying on filled prescription data to define exposure which, as demonstrated, can introduce exposure misclassification and underestimate the actual risks [19].

16

In conclusion, by studying a large number of exposed pregnancies, isolating the independent effect of ondansetron on risk, minimizing the risk of exposure misclassification and controlling for potential confounders, we have demonstrated significantly elevated risks for numerous structural

SC RI PT

defects. Birth defects carry a significant burden on individuals, public health, and the health care system. They are associated with increased risk of mortality [44] and many are associated with life-long disabilities and chronic health issues. It is incumbent upon public health professionals and medical practitioners to use the best available data to inform policy and practice in ways that improve health outcomes and quality of life for the patients and populations we serve.

Drug

U

exposures represent a modifiable risk factor, particularly for those scenarios in which alternative

N

therapies exist. Our study provides evidence that first trimester ondansetron exposure is associated

A

with increased risk of various structural birth defects and sheds important light on the magnitude

M

of exposure misclassification that may have obscured this association in prior studies using

CONFLICT OF INTEREST

TE

5

D

prescription data.

EP

The authors declare that there was no outside involvement in study design; in the collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit

CC

the manuscript for publication.

A

As an organization TTi reports receiving funds from plaintiff law firms involved in ondansetron litigation and a manufacturer of ondansetron.

17

6

ACKNOWLEDGEMENTS

The authors would like to thank Case Zambelli, Robert Bauserman, and Carter Little for their

A

CC

EP

TE

D

M

A

N

U

SC RI PT

assistance with manuscript preparation and review.

18

REFERENCES

A

CC

EP

TE

D

M

A

N

U

SC RI PT

[1] N.M. Lee, S. Saha, Nausea and Vomiting of Pregnancy, Gastroenterol. Clin. North Am. 40 (2011) 309–334. doi:10.1016/j.gtc.2011.03.009. [2] T.R. Einarson, C. Piwko, G. Koren, Quantifying the global rates of nausea and vomiting of pregnancy: a meta analysis, J. Popul. Ther. Clin. Pharmacol. J. Ther. Popul. Pharamcologie Clin. 20 (2013) e171-183. [3] L. Fiaschi, C. Nelson-Piercy, L.J. Tata, Hospital admission for hyperemesis gravidarum: a nationwide study of occurrence, reoccurrence and risk factors among 8.2 million pregnancies, Hum. Reprod. 31 (2016) 1675–1684. doi:10.1093/humrep/dew128. [4] M.F.G. Verberg, D.J. Gillott, N. Al-Fardan, J.G. Grudzinskas, Hyperemesis gravidarum, a literature review, Hum. Reprod. Update. 11 (2005) 527–539. doi:10.1093/humupd/dmi021. [5] F. Borrelli, R. Capasso, G. Aviello, M.H. Pittler, A.A. Izzo, Effectiveness and Safety of Ginger in the Treatment of Pregnancy-Induced Nausea and Vomiting, Obstet. Gynecol. 105 (2005) 849. doi:10.1097/01.AOG.0000154890.47642.23. [6] N. Ebrahimi, C. Maltepe, A. Einarson, Optimal management of nausea and vomiting of pregnancy, Int. J. Womens Health. 2 (2010) 241–248. [7] G. Ozgoli, M. Goli, M. Simbar, Effects of Ginger Capsules on Pregnancy, Nausea, and Vomiting, J. Altern. Complement. Med. 15 (2009) 243–246. doi:10.1089/acm.2008.0406. [8] R. Bishai, P. Mazzotta, G. Atanackovic, Z. Levichek, M. Pole, L.A. Magee, G. Koren, Critical appraisal of drug therapy for nausea and vomiting of pregnancy: II. Efficacy and safety of diclectin (doxylamine-B6)., Can. J. Clin. Pharmacol. J. Can. Pharmacol. Clin. 7 (2000) 138–143. [9] L.A. Magee, P. Mazzotta, G. Koren, Evidence-based view of safety and effectiveness of pharmacologic therapy for nausea and vomiting of pregnancy (NVP), Am. J. Obstet. Gynecol. 186 (2002) S256–S261. doi:10.1067/mob.2002.122596. [10] American College of Obstetrics and Gynecology, ACOG (American College of Obstetrics and Gynecology) Practice Bulletin: nausea and vomiting of pregnancy, Obstet. Gynecol. 103 (2004) 803–814. [11] G. Koren, S. Clark, G.D.V. Hankins, S.N. Caritis, J.G. Umans, M. Miodovnik, D.R. Mattison, I. Matok, Maternal safety of the delayed-release doxylamine and pyridoxine combination for nausea and vomiting of pregnancy; a randomized placebo controlled trial, BMC Pregnancy Childbirth. 15 (2015). doi:10.1186/s12884-015-0488-1. [12] L.G. Taylor, S.T. Bird, L. Sahin, M.S. Tassinari, P. Greene, M.E. Reichman, S.E. Andrade, K. Haffenreffer, S. Toh, Antiemetic use among pregnant women in the United States: the escalating use of ondansetron, Pharmacoepidemiol. Drug Saf. 26 (2017) 592–596. doi:10.1002/pds.4185. [13] Committee on Practice Bulletins-Obstetrics, ACOG Practice Bulletin No. 189: Nausea And Vomiting Of Pregnancy, Obstet. Gynecol. 131 (2018) e15–e30. doi:10.1097/AOG.0000000000002456. [14] S.E. Parker, C. Van Bennekom, M. Anderka, A.A. Mitchell, National Birth Defects Prevention Study, Ondansetron for Treatment of Nausea and Vomiting of Pregnancy and the Risk of Specific Birth Defects, Obstet. Gynecol. 132 (2018) 385–394. doi:10.1097/AOG.0000000000002679.

19

A

CC

EP

TE

D

M

A

N

U

SC RI PT

[15] B. Danielsson, B.N. Wikner, B. Källén, Use of ondansetron during pregnancy and congenital malformations in the infant, Reprod. Toxicol. 50 (2014) 134–137. doi:10.1016/j.reprotox.2014.10.017. [16] M. Anderka, A.A. Mitchell, C. Louik, M.M. Werler, S. Hernández-Diaz, S.A. Rasmussen, the National Birth Defects Prevention Study, Medications used to treat nausea and vomiting of pregnancy and the risk of selected birth defects, Birt. Defects Res. A. Clin. Mol. Teratol. 94 (2012) 22–30. doi:10.1002/bdra.22865. [17] C. Asker, B.N. Wikner, B. Källén, Use of antiemetic drugs during pregnancy in Sweden, Eur. J. Clin. Pharmacol. 61 (2005) 899. doi:10.1007/s00228-005-0055-1. [18] L. Colvin, A.W. Gill, L. Slack-Smith, F.J. Stanley, C. Bower, Off-Label Use of Ondansetron in Pregnancy in Western Australia, BioMed Res. Int. (2013). doi:10.1155/2013/909860. [19] B. Pasternak, H. Svanström, A. Hviid, Ondansetron in Pregnancy and Risk of Adverse Fetal Outcomes, N. Engl. J. Med. 368 (2013) 814–823. doi:10.1056/NEJMoa1211035. [20] Z. Shahraki, Z.S.H. Bonjar, F. Forghani, R. Nakhai, Comparing Neonatal Outcome Following the Use of Ondansetron versus Vitamin B6 in Pregnant Females with Morning Sickness: A Randomized Clinical Trial, J. Compr. Pediatr. 7 (2016). doi:10.17795/compreped-37081. [21] A. Einarson, C. Maltepe, Y. Navioz, D. Kennedy, M.P. Tan, G. Koren, The safety of ondansetron for nausea and vomiting of pregnancy: a prospective comparative study, BJOG Int. J. Obstet. Gynaecol. 111 (2004) 940–943. doi:10.1111/j.1471-0528.2004.00236.x. [22] Andersen, Jon, Jimenez-Solem, Poulsen, Henrik, Ondansetron use in early pregnancy and the risk of congenital malformations, at International Society of Pharmacoepidemiology: 2013. [23] M.M. Werler, M.M. Yazdy, J.R. Kasser, S.T. Mahan, R.E. Meyer, M. Anderka, C.M. Druschel, A.A. Mitchell, Medication Use in Pregnancy in Relation to the Risk of Isolated Clubfoot in Offspring, Am. J. Epidemiol. 180 (2014) 86–93. doi:10.1093/aje/kwu096. [24] Van Bennekom et al., Ondansetron for teh treatment of nausea and vomiting of pregnancy and the risk of birth defects, in: Boston, MA, 2015. [25] M.J. Funk, S.N. Landi, Misclassification in Administrative Claims Data: Quantifying the Impact on Treatment Effect Estimates, Curr. Epidemiol. Rep. 1 (2014) 175–185. doi:10.1007/s40471-014-0027-z. [26] L. Hansen, IBM MarketScan Research Databases for life sciences researchers., (n.d.). https://www-01.ibm.com/common/ssi/cgi-bin/ssialias?htmlfid=HLW03049USEN. [27] J.A. Cole, S.A. Ephross, I.S. Cosmatos, A.M. Walker, Paroxetine in the first trimester and the prevalence of congenital malformations, Pharmacoepidemiol. Drug Saf. 16 (2007) 1075–1085. doi:10.1002/pds.1463. [28] G. Maldonado, S. Greenland, Simulation study of confounder-selection strategies, Am. J. Epidemiol. 138 (1993) 923–936. [29] K.J. Rothman, S. Greenland, T.L. Lash, Modern Epidemiology, Lippincott Williams & Wilkins, 2008. [30] G. Koren, S. Madjunkova, C. Maltepe, The protective effects of nausea and vomiting of pregnancy against adverse fetal outcome--a systematic review, Reprod. Toxicol. Elmsford N. 47 (2014) 77–80. doi:10.1016/j.reprotox.2014.05.012. [31] R.M. Weigel, M.M. Weigel, Nausea and vomiting of early pregnancy and pregnancy outcome. A meta-analytical review, Br. J. Obstet. Gynaecol. 96 (1989) 1312–1318. 20

A

CC

EP

TE

D

M

A

N

U

SC RI PT

[32] S.N. Hinkle, S.L. Mumford, K.L. Grantz, R.M. Silver, E.M. Mitchell, L.A. Sjaarda, R.G. Radin, N.J. Perkins, N. Galai, E.F. Schisterman, Association of Nausea and Vomiting During Pregnancy With Pregnancy Loss: A Secondary Analysis of a Randomized Clinical Trial, JAMA Intern. Med. 176 (2016) 1621–1627. doi:10.1001/jamainternmed.2016.5641. [33] E.A. Mayhall, R. Gray, V. Lopes, K.A. Matteson, Comparison of antiemetics for nausea and vomiting of pregnancy in an emergency department setting, Am. J. Emerg. Med. 33 (2015) 882–886. doi:10.1016/j.ajem.2015.03.032. [34] S.-S.N. Siu, M.T.V. Chan, T.-K. Lau, Placental Transfer of Ondansetron during Early Human Pregnancy, Clin. Pharmacokinet. 45 (2006) 419–423. doi:10.2165/00003088200645040-00006. [35] C. for D.E. and Research, Drug Safety and Availability - FDA Drug Safety Communication: Abnormal heart rhythms may be associated with use of Zofran (ondansetron), (n.d.). https://www.fda.gov/Drugs/DrugSafety/ucm271913.htm (accessed December 16, 2017). [36] S. Greenland, Basic Methods for Sensitivity Analysis of Biases, Int. J. Epidemiol. 25 (1996) 1107–1116. doi:10.1093/ije/25.6.1107-a. [37] J.A. Smith, J.S. Refuerzo, K.A. Fox, Treatment and outcome of nausia and vomiting of pregnancy, UpToDate. (2018). [38] K.F. Huybrechts, G. Bröms, L.B. Christensen, K. Einarsdóttir, A. Engeland, K. Furu, M. Gissler, S. Hernandez-Diaz, P. Karlsson, Ø. Karlstad, H. Kieler, A.-M. LahesmaaKorpinen, H. Mogun, M. Nørgaard, J. Reutfors, H.T. Sørensen, H. Zoega, B.T. Bateman, Association Between Methylphenidate and Amphetamine Use in Pregnancy and Risk of Congenital Malformations: A Cohort Study From the International Pregnancy Safety Study Consortium, JAMA Psychiatry. 75 (2018) 167–175. doi:10.1001/jamapsychiatry.2017.3644. [39] B.K. Frohnert, R.C. Lussky, M.A. Alms, N.J. Mendelsohn, D.M. Symonik, M.C. Falken, Validity of Hospital Discharge Data for Identifying Infants with Cardiac Defects, J. Perinatol. 25 (2005) 737–742. doi:10.1038/sj.jp.7211382. [40] S.D. Carstairs, Ondansetron Use in Pregnancy and Birth Defects: A Systematic Review, Obstet. Gynecol. 127 (2016) 878–883. doi:10.1097/AOG.0000000000001388. [41] M.S. Fejzo, K.W. MacGibbon, P.M. Mullin, Ondansetron in pregnancy and risk of adverse fetal outcomes in the United States, Reprod. Toxicol. 62 (2016) 87–91. doi:10.1016/j.reprotox.2016.04.027. [42] Ş. Özdemirci, F. Akpınar, M. Bilge, F. Özdemirci, S. Yılmaz, D. Esinler, İ. Kahyaoğlu, The Safety of Ondansetron and Chlorpromazine for Hyperemesis Gravidarum in First Trimester Pregnancy, Gynecol. Obstet. Reprod. Med. 20 (2016). http://gorm.com.tr/index.php/GORM/article/view/140 (accessed January 10, 2018). [43] A. Bérard, K.L. Wisner, S. Hultzsch, C. Chambers, Field studies versus database studies on the risks and benefits of medication use during pregnancy: Distinct pieces of the same puzzle, Reprod. Toxicol. 60 (2016) 123–128. doi:10.1016/j.reprotox.2016.02.002. [44] A. Correa-Villaseñor, J. Cragan, J. Kucik, L. O’Leary, C. Siffel, L. Williams, The Metropolitan Atlanta Congenital Defects Program: 35 years of birth defects surveillance at the Centers for Disease Control and Prevention, Birt. Defects Res. A. Clin. Mol. Teratol. 67 (2003) 617–624. doi:10.1002/bdra.10111.

21

N U SC RI PT

Table 1. Characteristics of Women Included in the Analysis, According to Birth Outcome Variable

M

A

Maternal age at birth (mean ± SD) Year of birth (mean ± SD) Gender of baby Male Female

A

CC E

PT

ED

Unknown/unspecified Region of birth Northeast North Central South West Unknown Medical history - no. (%) Obesity Diabetes mellitus Epilepsy Hypertension Cancer diagnosed Prior preterm delivery Family history of birth defects Ondansetron exposure - no. (%)a Medical administration (only) Prescription (any)

No Birth Defects n=802,253

Cardiac Birth Defects n=32,100

Orofacial Cleft Birth Defects n=1,590

31.63 (4.59) 2009 (3.25)

32.22 (4.72) 2010 (3.01)

31.83 (4.68) 2010 (3.30)

385,594 (48.06) 386,511 (48.18) 30,148 (3.76)

16,050 (50.00) 15,087 (47.00) 963 (3.00)

875 (55.03) 661 (41.57) 54(3.40)

141,381 (17.62) 206,129 (25.69) 301,428 (37.57) 140,779 (17.55) 12,536 (1.56)

5,599 (17.44) 7,758 (24.17) 13,274 (41.35) 5,058 (15.76) 411 (1.28)

273 (17.17) 428 (26.92) 620 (38.99) 246 (15.47) 23 (1.45)

28,392 (3.54) 13,695 (1.71) 2,464 (0.31) 23,430 (2.92) 6,406 (0.80) 9,471 (1.18) 3,158 (0.39)

1,778 (5.54) 1,184 (3.69) 137 (0.43) 1,714 (5.34) 327 (1.02) 601 (1.87) 227 (0.71)

70 (4.40) 49 (3.08) 6 (0.38) 65 (4.09) 17 (1.07) 27 (1.70) 19 (1.19)

5,042 (0.68) 70,215 (8.75)

303 (1.03) 3,099 (9.65)

13 (0.90) 157 (9.87) 22

N U SC RI PT

Medications in early pregnancy- no. (%)

ED

M

A

Initiation of acid reducing therapy Psychotropics Prescription folic acid Macrolide antibiotics Corticosteroids Anticonvulsants Birth Outcomes - no. (%) Low Birth Weight Multiple Gestation High risk pregnancy - no. (%)

PT

Diagnosis of nausea and vomiting in pregnancy or hyperemesis - no. (%)b

A

CC E

Nausea and vomiting during pregnancy Hyperemesis gravidarum Severe hyperemesis gravidarum

6,883 (0.86) 18,781 (2.34) 89,846 (11.20) 79,262 (9.88) 33,680 (4.20) 8,233 (1.03)

289 (0.90) 1,051 (3.27) 3,482 (10.85) 3,461 (10.78) 1,375 (4.28) 446 (1.39)

22 (1.38) 51 (3.21) 198 (12.45) 178 (11.19) 74 (4.65) 31 (1.95)

13,750 (1.71) 13,750 (1.71) 9,479 (1.18)

2,644 (8.24) 1,277 (3.98) 469 (1.46)

89 (5.60) 41 (2.58) 27 (1.70)

46,559 (5.80)

2,263 (7.05)

90 (5.66)

1,834 (0.23) 533 (0.07)

61 (0.19) 19 (0.06)

1 (0.06) 0 (0.00)

126 (0.02)

6 (0.02)

0 (0.00)

Observed differences between the no birth defect group and the birth defect groups were statistically significant at p<0.05 except for region, initiation of acid-reducing therapy, prescription folic acid use, and NVP and HG. a. Exposed mother-child pairs were exclusively exposed to ondansetron during first trimester, meaning there was not a claim for Diclegis®, metoclopramide, promethazine or methylprednisolone during the first trimester period. b. Any diagnosis of NVP, HG, severe HG, or late-vomiting anytime during pregnancy.

23

N U SC RI PT

Table 2. Association of Ondansetron Exposure with Structural Birth Defects Exposed in First Trimester (Rx or Med) n=76,330

Unadjusted

Adjusteda

29,001 (3.67) 1,433 (0.18)

303 (5.45) 13 (0.23)

3,099 (4.06) 157 (0.21)

1.52 (1.35-1.70) 1.32 (0.76-2.28)

1.43 (1.28-1.61) 1.30 (0.75-2.25)

1.11 (1.07-1.16) 1.14 (0.97-1.35)

1.04 (1.00-1.08) 1.12 (0.95-1.33)

301 (5.42) 81 (1.46)

3,069 (4.02) 883 (1.16)

1.53 (1.36-1.71) 1.30 (1.04-1.62)

1.44 (1.28-1.62) 1.29 (1.03-1.61)

1.12 (1.08-1.16) 1.02 (0.95-1.09)

1.04 (1.00-1.08) 1.00 (0.93-1.07)

23,903 (3.03) 813 (0.10)

267 (4.81) 15 (0.27)

2,633 (3.45) 97 (0.13)

1.62 (1.43-1.84) 2.68 (1.61-4.47)

1.49 (1.32-1.69) 2.71 (1.62-4.52)

1.15 (1.10-1.20) 1.24 (1.01-1.54)

1.04 (0.99-1.08) 1.24 (1.00-1.53)

343 (0.04)

5 (0.09)

43 (0.06)

2.12 (0.88-5.12)

2.12 (0.87-5.13)

1.31 (0.95-1.80)

1.31 (0.95-1.81)

6,044 (0.77)

76 (1.37)

678 (0.89)

1.83 (1.45-2.29)

1.75 (1.39-2.20)

1.17 (1.08-1.27)

1.11 (1.02-1.20)

1,433 (0.18) 1,068 (0.14) 638 (0.08) 696 (0.09)

13 (0.23) 11 (0.20) 5 (0.09) 8 (0.14)

157 (0.21) 112 (0.15) 74 (0.10) 71 (0.09)

1.32 (0.76-2.28) 1.49 (0.53-2.71) 1.14 (0.47-2.74) 1.67 (0.83-3.35)

1.30 (0.75-2.25) 1.46 (0.81-2.65) 1.16 (0.48-2.81) 1.69 (0.84-3.40)

1.14 (0.97-1.35) 1.09 (0.90-1.33) 1.21 (0.95-1.54) 1.06 (0.83-1.36)

1.12 (0.95-1.33) 1.06 (0.87-1.30) 1.22 (0.96-1.56) 1.07 (0.84-1.38)

6,295 (0.80) 11,7559 (1.49) 409 (0.05) 7,803 (0.99)

56 (1.01) 94 (1.69) 7 (0.13) 69 (1.24)

808 (1.06) 1,329 (1.74) 55 (0.07) 835 (1.09)

1.29 (0.99-1.68) 1.16 (0.95-1.43) 2.49 (1.18-5.25) 1.28 (1.01-1.63)

1.16 (0.89-1.51) 1.11 (0.90-1.36) 2.51 (1.19-5.31) 1.24 (0.98-1.58)

1.34 (1.24-1.44) 1.18 (1.11-1.25) 1.40 (1.06-1.86) 1.34 (1.24-1.44)

1.18 (1.09-1.27) 1.12 (1.05-1.18) 1.40 (1.05-1.87) 1.07 (1.00-1.16)

557 (0.07)

4 (0.07)

49 (0.06)

1.04 (0.39-2.79)

1.03 (0.38-2.75)

0.92 (0.69-1.23)

0.91 (0.67-1.22)

28,666 (3.63) 9,058 (1.15)

CC E

PT

ED

Primary Analysis Cardiac defects Orofacial clefting Secondary Analysis Septal defects Ventricular septal defect Atrial septal defect Atrioventricular septal defect Hypoplastic Left Heart Syndrome Other circulatory defects

n=787,753

Exposed in First Trimester (Med Only) n=5,557

A

Orofacial clefting Cleft palate Cleft lip Cleft lip with or without palate Laryngeal cleft Craniosynostosis Diaphragmatic hernia Renal collecting system anomalies Limb reduction defects

A

Unexposed in During Pregnancy

M

Outcome (no. %)

Prevalence Odds Ratio (95% CI) Medical Administration Only

24

Prevalence Odds Ratio (95% CI) Prescription or Medical Administration Unadjusted Adjusteda

113,106 (0.14)

N U SC RI PT

Other defects (negative control)

878 (0.16)

12,216 (0.16)

1.10 (1.03-1.18)

1.02 (0.95-1.10)

A

CC E

PT

ED

M

A

a. Odds ratios are adjusted for: mother's age, infant year of birth, and infant gender.

25

1.12 (1.09-1.14)

1.02 (1.00-1.04)

N U SC RI PT

Table 3. Sensitivity Analyses of Ondansetron Exposure in First trimester and Risk of Cardiac and Orofacial Birth Defects

1.47 (0.80-2.71)

1.48 (1.14-1.92)

1.43 (0.78-2.64)

NAb

1.54 (0.79-3.01)

NAb

1.29 (1.09-1.52)

1.30 (0.62-2.74)

1.30 (1.10-1.54)

1.31 (0.62-2.76)

1.56 (1.32-1.85)

1.38 (0.62-3.10)

1.60 (1.35-1.89)

1.41 (0.63-3.17)

M

1.56 (1.21-2.03)

Adjusted Analysis Cardiac Defects Orofacial Defects Prevalence Odds Prevalence Odds Ratio (95% CI) Medical Ratio (95% CI) Medical Administration Only Administration Only

1.66 (0.85-3.24)

PT

Control for Confounding by Indicationa Temporal Trends Year of Conception: 2000-2006 Year of Conception: 2007-2011 Year of Conception: 2012-2014

A

Unadjusted Analysis Cardiac Defects Orofacial Defects Prevalence Odds Prevalence Odds Ratio (95% CI) Medical Ratio (95% CI) Medical Administration Only Administration Only

ED

Type of Analysis Outcome

A

CC E

a. Control group is restricted to mother-child pairs with no antiemetic exposure during pregnancy and a diagnosis of NVP or HG. b. Association not available because of a zero cell in the cross-tabulation of the predictor and outcome.

26