This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Informatics, is properly cited. The complete bibliographic information, a link to the original publication on https://medinform.jmir.org/, as well as this copyright and license information must be included.
Medication-wide association studies (MWAS) have been applied to assess the risk of individual prescription use and a wide range of health outcomes, including cancer, acute myocardial infarction, acute liver failure, acute renal failure, and upper gastrointestinal ulcers. Current literature on the use of preconception and periconception medication and its association with the risk of multiple gestation pregnancies (eg, monozygotic and dizygotic) is largely based on assisted reproductive technology (ART) cohorts. However, among non-ART pregnancies, it is unknown whether other medications increase the risk of multifetal pregnancies.
This study aimed to investigate the risk of multiple gestational births (eg
We used electronic health record data between 2010 and 2017 on patients who delivered babies at Penn Medicine, a health care system in the Greater Philadelphia area. We explored 3 logistic regression models: model 1 (no adjustment); model 2 (adjustment for maternal age); and model 3—our final logistic regression model (adjustment for maternal age, ART use, and infertility diagnosis). In all models, multiple births (MBs) were our outcome of interest (binary outcome), and each medication was assessed separately as a binary variable. To assess our MWAS model performance, we defined ART medications as our gold standard, given that these medications are known to increase the risk of MB.
Of the 63,334 distinct deliveries in our cohort, only 1877 pregnancies (2.96%) were prescribed any medication during the preconception and first trimester period. Of the 123 medications prescribed, we found 26 (21.1%) medications associated with MB (using nominal
Our work demonstrates the opportunities in applying the MWAS approach with electronic health record data to explore associations between preconception and periconception medication exposure and the risk of MB while identifying novel candidate medications for further study. Overall, we found 3 novel medications linked with MB that could be explored in further work; this demonstrates the potential of our method to be used for hypothesis generation.
Multifetal pregnancies are at a high risk for obstetric complications, including anemia, preterm labor, pregnancy-induced hypertension, placental previa, and fetal malformations [
ART is a widely accepted treatment for infertile couples, referring to all treatments that include the handling of eggs, sperm, and embryos. Outside the scope of ART, hormonal medications for the purpose of facilitating a successful pregnancy are referred to as fertility treatment. Increased rates of monozygotic twinning have been observed in pregnancies due to ART use (ie, in vitro fertilization [IVF], micromanipulation, multiple embryo transfer, and gonadotrophin treatment) [
The wealth of information from electronic health record (EHR) data can allow for hypothesis-driven research on the associations between medications and pregnancy outcomes. Ryan et al [
Except for research conducted on nationwide health care data registries [
This study illustrates a proof-of-concept MWAS approach for hypothesis-driven pharmacovigilance research on EHR data, with a particular focus on MB.
We used EHR data obtained from 4 different hospitals within the Penn Medicine system: the Hospital of the University of Pennsylvania, the Pennsylvania Hospital, Penn Presbyterian Hospital, and Chester County Hospital. The deliveries were identified using a previously developed algorithm called Method to Acquire Delivery Date Information from Electronic Health Records (MADDIE) [
This study was approved by the institutional review board of the University of Pennsylvania (#828000).
Although a majority of twin births result from natural conception, the incidence of twins and other higher order multifetal pregnancy resulting from superovulation and ART is 20 times greater than the incidence from natural conception [
We mapped all inpatient and outpatient medications from Epic and other EHR systems to RxNorm using a previously described method [
We manually annotated the complete list of medications, adding the following elements: generic name, medication type, specific medication type, US Federal Drug Agency pregnancy category, associated comorbidities, and associations with pregnancy outcome treatment. We manually annotated this list because many drugs used in fertility treatments are used off-label; therefore, standardized medical terminology systems would be ineffective in capturing those use cases [
We constructed 3 logistic regression models with MB as our outcome of interest (binary outcome, 0 or 1), and the effect of each medication on the outcome was assessed separately (each medication exposure was a binary variable, coded 0 or 1). The analysis was performed using the general linear model function in R. The control group for each medication comprised all patients without exposure to the target medication (coded as 0), including patients who had no exposure to medications in the EHR data. Consequently, each target medication had its own control group. We adjusted for 3 known confounders of MB: maternal age (encounter age), ART-resulting pregnancy diagnoses (0 or 1), and infertility diagnoses (0 or 1;
Significant medications (
We categorized medications with significant nominal
We obtained EHR data from 1,060,100 female patients treated at Penn Medicine, with inpatient and outpatient visits between 2010 and 2017. A previously developed algorithm called MADDIE identified 50,560 patients who delivered a baby at Penn Medicine having 63,334 distinct deliveries [
Retrospective cohort selection process. MADDIE: Method to Acquire Delivery Date Information from Electronic Health Records; MWAS: medication-wide association study.
We manually annotated 123 medications that were prescribed during the preconception period and the first trimester period of 1877 pregnancies of the 63,334 (2.96%) total distinct deliveries in our cohort (
Aside from fertility medications, the list contained several types of pain (15/123, 12.2%), antibiotic (11/123, 8.9%), and antihistamine medications (8/123, 6.5%). Most of the extracted medications were not formally assigned (48/123, 39%), followed by category C (31/123, 25.2%) and category B (24/123, 19.5%) medications. As expected, fewer medications were categorized as category A (2/123, 1.6%) and category D (5/123, 4.1%). We found 9.8% (12/123) of medications were categorized as category X, contraindicated in pregnancy, medications—all of which are medications indicated for fertility treatment, contraception, or other indications in obstetrics and gynecology practice.
Retrospective cohort medication exposure data.
|
Total distinct deliveries (N=63,334) | No prescription medication exposure (n=61,457) | Prescription medication exposurea (n=1877) | Fertility medicationb exposure (n=231) | |
|
|||||
|
Multiple birthc | 1562 (2.47) | 1476 (2.4) | 86 (4.58) | 37 (16) |
|
|||||
|
Assisted reproductive technologyd | 246 (0.39) | 218 (0.35) | 28 (1.49) | 16 (6.9) |
|
Infertilitye | 48 (0.08) | 39 (0.06) | 9 (0.48) | 4 (1.7) |
Maternal age, mean (SD) | 29.5 (6.1) | 29.5 (6.1) | 30.5 (5.7) | 34.6 (4.0) |
aPrescription medication exposure is during the preconception period and the first trimester period only in this cohort.
b
cMultiple birth determined by International Classification of Diseases (ICD) codes shown in
dPregnancy resulting from assisted reproductive was determined by the ICD codes shown in
eInfertility diagnosis determined by ICD codes shown in
In
Medications and covariates significantly associated with multiple birth, using odds ratio (95% CIs). Medication names found significant in our logistic regression model 3 (
Validation set performance was evaluated for our fully adjusted model (model 3). Of the 123 medications extracted, we found 26 (21.1%) medications nominally associated with MB (
Performance validation of assisted reproductive technology medications in medication-wide association study.
|
Performance metrica | |||||
|
Sensitivity | Specificity | Accuracy | Precision | F1 score | |
|
||||||
|
.80 | .84 | .84 | .41 | .54 | |
|
.47 | .96 | .90 | .64 | .54 | |
|
||||||
|
.73 | .85 | .84 | .41 | .53 | |
|
.47 | .96 | .90 | .64 | .50 | |
|
||||||
|
.73 | .86 | .85 | .42 | .53 | |
|
.40 | .96 | .89 | .60 | .48 |
aPerformance metrics were calculated using formulas shown in
bMaternal age determined by age at delivery encounter.
cPregnancy resulting from assisted reproductive technology determined by the International Classification of Diseases codes shown in
dInfertility diagnosis determined by the International Classification of Diseases codes shown in
Prescription medications associated with comorbidities of fertility and ART treatment were found, as well as medications that may be used for obstetric complications related to multifetal pregnancy care. Of the 26 significant medications using nominal
Medications associated with multiple birth after adjustment for assisted reproductive technology, infertility, and maternal age (model 3).
Indicated comorbidity | Generic medication name or names | Medications associated with multiple birth, n (%) | |||
|
|||||
|
Assisted reproductive technology treatment | EMLA, methylprednisolone, diazepam, amoxicillin, doxycycline, and medroxyprogesterone acetate | 6 (23) | ||
|
Asthma | Albuterol, fluticasone propionate and salmeterol, and epinephrine | 3 (12) | ||
|
Irritable bowel disease | Dicyclomine | 1 (4) | ||
|
|||||
|
Cardiovascular-related diagnoses (gestational hypertension and thrombosis) | Heparin | 1 (4) | ||
|
Gestational diabetes mellitus | Insulin aspart, human | 1 (4) | ||
Not previously associated with multiple birth, assisted reproductive technology, or fertility-related problems | Sumatriptan, oxytocin, and lorazepam | 3 (12) |
Conceptual schema for medication-wide association study (MWAS) analyses on multiple birth. Confounding relationships for medication-outcome associations are illustrated. Within the MWAS, we adjust for maternal age, infertility diagnosis, and assisted reproductive technology–resulting pregnancy diagnosis. The study does not adjust for all known associations of multiple birth such as obstetric complications or family history of multiples. The validation of the MWAS models observed performance in capturing fertility medication exposure.
We applied 3 logistic regression models to retrospective EHR data of a cohort of patients who delivered at Penn Medicine between 2010 and 2017 (n=63,334) to explore potential associations between the medications prescribed during the preconception and first trimester period (binary variable) and the occurrence of MB (binary outcome). We discuss the results of our MWAS from our fully adjusted model that was adjusted for age and ART and infertility diagnosis (model 3) on MB for all associations revealed using nominal.
The application of an MWAS approach to MB allows the analysis of medications used outside the scope of obstetric treatment, capturing comorbidities that may increase the risk of the outcome. Not all of this is known, as MB is more commonly used as an adjustment for analysis of other pregnancy outcomes of interest. Off-label use is common in pregnancy and infertility treatment [
The graph in
MB outcomes can also be observed in ART cohort databases, such as the SART Clinical Outcomes Reporting System. However, fertility medication treatment without the intention of egg retrieval will not necessarily be captured within such databases, as they are beyond the scope of ART. Moreover, not all multifetal pregnancies result from infertility treatment and ART. Finally, such databases are reported from ART clinics and are not necessarily representative of all the medications prescribed during pregnancy. An ART cohort database may have a wealth of data elements specific to ART treatment; however, these data are reported using inconsistent methods, often from a variety of reporting services [
We found that 5.51% (86/1562) MB pregnancies had prescription medication exposure during pregnancy. Therefore, pregnancies resulting in MB were more likely to have recorded prescription medication during the preconception and first trimester period. This is consistent with (1) the fact that ART often uses medications early on to induce pregnancy [
To assess the ability of our MWAS to capture medications that increase the risk of MB, we used medications that are known to increase an individual’s chance of conceiving and have been implicated in increasing the risk of MB in the literature (
Medications used in fertility treatment themselves may be captured solely because of reverse causation, although they do not have a truly strong association with multifetal pregnancy. Several medications may be prescribed during IVF treatment cycles for preventive care or other indications, including antibiotics (doxycycline and amoxicillin), a corticosteroid (methylprednisolone), pain management (EMLA), progestin-induced menstruation (medroxyprogesterone acetate), and conscious sedation (diazepam) [
ART and ovulation induction procedures are used for fertility treatment. A comprehensive review of infertility comorbidities in women suggests that infertility is a complex health care issue, and women with infertility are at a higher risk of psychiatric disorders and endometrial cancer [
Medications associated with comorbidities of infertility were identified, including treatments for asthma and IBD. Research shows that women with asthma have higher pregnancy losses [
Medications identified by the MWAS may be prescribed for obstetric complications associated with multifetal pregnancy. These pregnancies are at an increased risk of obstetric complications, such as preterm birth, placental problems, gestational diabetes mellitus, anemia, and preeclampsia. Owing to the time exposure range, medications typically used to treat complications typically past the first trimester of pregnancy were not captured by the MWAS. One antidiabetic medication (ie, insulin aspart, human) was identified; however, other forms of insulin and the insulin sensitizer metformin were not identified as significant. A single antithrombotic medication, heparin, was identified, but other anticoagulants and cardiovascular-related medications were not identified in our models.
Migraines have a high incidence in obstetrics; one migraine pharmacological treatment (sumatriptan) was found to be associated with MB. An association between migraine history and development of ovarian hyperstimulation syndrome may indicate the risk of multifetal pregnancy [
Although the World Health Organization’s anatomical therapeutic chemical classification system is applicable, the proportion of RxNorm drugs mapped to anatomical therapeutic chemicals would result in fewer medications being included in the analysis. However, the therapeutic use of the medications has not been explicitly determined. Medications classified as fertility related are based on the SART references; however, without discrete indications, they potentially underpower performance in the validation process. In addition, a major limitation of using standard pharmacology and drug-related terminologies is that approximately 11% of medications used in women’s health are off-label [
Our research demonstrates opportunities in using an MWAS approach with EHR data to explore agents previously unknown to be associated with MB outcomes. The results indicated that a number of medications used in ART and infertility treatment were associated with an increased incidence of MB likely due to multifetal pregnancy, as expected. Using these medications as our gold standard, we found that our algorithm had an accuracy of 85% and 89%, using nominal
International Classification of Diseases (ICD)-9 and ICD-10 codes to capture multiple birth.
International Classification of Diseases (ICD)-9 and ICD-10 codes used to identify comorbidity diagnosis from the electronic health record for model adjustment.
Formulas for validation.
Medications that can be indicated for infertility treatment.
Medication-wide association study results for models 1, 2, and 3.
Odds ratio with 95% CIs for observed prescription of medications.
Performance analysis confusion tables.
Graph of patients with assisted reproductive technology or fertility diagnosis and patients with fertility medication prescriptions.
assisted reproductive technology
electronic health record
irritable bowel disease
International Classification of Diseases
in vitro fertilization
Method to Acquire Delivery Date Information from Electronic Health Records
multiple birth
medication-wide association study
odds ratio
Society for Assisted Reproductive Technology
The authors would like to thank the Perelman School of Medicine at the University of Pennsylvania for providing generous funds to support this project.
None declared.