Prescription Drugs in Pregnancy and Congenital Malformations: A Population‐Based Safety Screening Study

ABSTRACT

Background

Counseling women who are either pregnant or contemplating pregnancy on their use of prescribed drugs remains a major clinical challenge. Since the thalidomide tragedy in the 1960s, the use of drugs during pregnancy has been subject to widespread concern due to the potential for unwanted effects on the unborn child, notably major congenital malformations.

Objective

To examine the risk of major congenital malformations following first trimester exposure to all marketed prescription drugs in Denmark.

Study Design

This was a population‐based cohort study utilizing national health registries in Denmark. We studied all singleton livebirths in Denmark between January 1, 2004, and December 31, 2017, and we linked data from the National Danish Prescription Register, Birth Register, Patient Register, and Cause of Death Register. Using logistic regression analysis, we compared exposed liveborn to unexposed liveborn children while controlling for important confounders. The main outcome measure was major congenital malformations, and the secondary outcomes included organ‐specific major congenital malformations as defined by EUROCAT.

Results

Of 326 drugs with at least 5 livebirths with major congenital malformations, 31 were associated with an increased risk of major congenital malformations compared to unexposed livebirths (adjusted Odds Ratio [aOR] ≥ 2.0). Compared to livebirths of women who discontinued treatment prior to pregnancy, 17 drugs with an increased risk (aOR ≥ 2.0) were identified. Among 115 drugs prescribed to ≥ 1000 women during the first trimester, only insulins had aORs ≥ 2.0 for overall major congenital malformations. There were > 100 drugs with no increased risk of major congenital malformations.

Conclusions

Using a complete nationwide dataset, > 100 null‐associations between first‐trimester drug exposure and overall major congenital malformations were documented. This provides important insights and reassurance to support pregnant women and inform shared decision making. We confirm previously known teratogenic drugs and other potential teratogenic drugs, clopidogrel and liraglutide, were identified. These latter associations should be addressed in future studies using disease‐specific confounder control.

Summary.

• Why was this study conducted?

  • ○
    Use of drugs in pregnancy and risk of congenital malformations remain a challenge to decision making in everyday clinical practice. We performed a nation‐wide registry‐based screening of all prescription drugs used in first trimester pregnancy and the ensuing risk of congenital malformations among live‐born children.
• What are the key findings?

  • ○
    We documented no risk for 112 drugs with at least 1000 live‐born exposed children; confirmed known teratogens and identified new potential teratogens for future specific studies.
• What does this study add to what is already known?

  • ○
    This is the first complete nation‐wide set of such data and provide a solid hypothesis generating starting point for future specific studies. Our results provide substantial evidence to decision support not least with many reassuring null‐associations.

1. Introduction

Counseling women who are either pregnant or contemplating pregnancy on their use of prescribed drugs is a major clinical challenge. Since the thalidomide tragedy in the 1960s, the use of drugs during pregnancy has been subject to widespread concern due to potential for unwanted effect to the unborn child, notably major congenital malformations (MCM) [1, 2, 3, 4]. Many pregnant women discontinue treatment during pregnancy, occasionally due to unsubstantiated fear of drug‐induced MCM, or on the advice of their treating physician [5, 6, 7]. However, certain maternal conditions cannot be left untreated, for example, diabetes, major depression, hypertension, or epilepsy, leaving women and their physicians with difficult decisions on medical treatment.

MCM are an important cause of fetal death, infant mortality and morbidity, and each year approximately 2.5%–3.0% of livebirths in the European Union are born with a MCM according to European network of population‐based registries for the epidemiological surveillance of congenital anomalies (EUROCAT) estimates [8, 9]. When a new drug is awarded marketing authorization, little or no information is available on the safety when used in pregnant women as they are routinely excluded from clinical trials. Observational studies are essential to collect evidence‐based information on adverse effects to the fetus, notably risks of MCM.

Using the comprehensive and nationwide administrative health care registers in Denmark, the objective was to examine the risk of major congenital malformations among liveborn children following first trimester exposure to all marketed prescription drugs in Denmark.

2. Materials and Methods

We conducted a cohort study based on all livebirths in Denmark between January 1, 2004, and December 31, 2017. We studied all individual prescription drugs used in first trimester pregnancy and compared the risk of MCM against unexposed livebirths in our primary analysis and against livebirths to mothers who discontinued treatment with the specific drug prior to pregnancy as a sensitivity analysis. Registry‐based studies does not require Institutional Review Board approval at the University of Southern Denmark. In accordance with Danish law, ethical approval and informed consent is not required for register‐based studies. The project was approved by the Danish Data Protection Agency, journal number 19/36067.

2.1. Setting and Data Sources

All Danish residents (population 5.9 million) have access to government‐funded health care, and all hospital contacts are recorded for administrative purposes in comprehensive health registries. The study population was identified using the Danish Medical Birth registry [10] (the Birth register). Using a unique person identifier from the Danish Civil Registration Registry [11], we linked all identified women to prescription drug data from the Danish National Prescription Registry [12] (the Prescription register), hospital records for both women and their offspring using the Danish National Patient registry [13] (the Patient register), and cause of death from the Cause of Death Register [14].

The Birth register provided the biological mother‐baby link in addition to a broad variety of data on the mother, complications during pregnancy and delivery, details on the infant at delivery including both stillbirths after 22 weeks of gestation and livebirths at any gestational age, and information on diagnosed malformations or date of death within the first postnatal year. There are about 60 000 livebirths in Denmark (www.dst.dk) every year.

The Prescription register contains individual‐level prescription data from all Danish outpatient pharmacies. Each record contains data on recipient identification number, date of collection, generic drug prescribed, number of tablets/units, strength, and amount of defined daily doses (DDD).

2.2. Study Population

We identified all singleton livebirths in Denmark between January 1, 2004, and December 31, 2017, while allowing for 1‐year of follow‐up (December 31, 2018). Due to substantial restructuring of the Danish National Registries in 2019, and an expected transition period with uncertain completeness of data, the study period was not extended beyond 2018.

We excluded pregnancies with missing or implausible gestational age records and women not residing in Denmark for at least 1 year before delivery to improve classification of timing of exposure and complete data coverage. Implausible gestational age was defined by recorded gestational age less than 23 completed weeks if the livebirth survived more than a fortnight. Infants with chromosomal abnormalities (ICD‐10 codes Q90–Q99) were excluded from all analyses. We excluded women who never redeemed any prescription in their life as they likely represent an overall “healthier” cohort which may potentially dilute the non‐exposed comparator group. The non‐exposed comparator group comprises women who did not redeem any prescription during or 1 year prior to pregnancy.

2.3. Classification of Exposure

Exposure defined by fifth level ATC classification was evaluated separately. Livebirths were considered exposed if their mother had filled a prescription between the first day in the last menstrual period (LMP) and the end of the first trimester (91 days after LMP). LMP is defined by the day of birth minus gestational age, which for > 95% was determined at first trimester scanning. A filled prescription within 30 days prior to LMP, with a duration of treatment overlapping (assuming a daily intake of one DDD) LMP was also considered as defining exposure. Unexposed livebirths were defined by no drug (any drug) exposure between 90 days prior to LMP and the end of the first trimester. The use of more than one drug during pregnancy was not addressed. To mitigate confounding by indication, second comparator cohorts of unexposed livebirths were established for each individual drug. This comprised a group of livebirths born to women who were treated with the individual drug of interest during the last year before pregnancy, but not during pregnancy (Figure 1). These three cohorts were mutually exclusive.

FIGURE 1.

Definition of exposure. LMP: First day of the last menstrual period. Exposed: At least one filled prescription for the specific drug in question.

2.4. Core Outcome Sets

MCM were identified at the time of birth or within 365 days of delivery to ensure inclusion of late detected MCM, which is of importance for some cardiac malformations [15]. MCM were identified using infant records from three Danish health registries: The Birth register [10], the Patient register [13], and the Cause of Death Register [14] during the first year of life. All contact and discharge diagnoses were categorized using the 10th version of the International Classification of Diseases (ICD‐10).

Overall and organ specific MCM were identified using a Danish adaptation of the EUROCAT classification of malformations (previously described elsewhere) [16, 17]. Primary and secondary diagnoses were included, and major congenital hip malformations (Q65.0‐2) were restricted to those with a second such diagnosis 6 weeks post‐partum to exclude records of tentative diagnosis referred for further diagnostic tests (i.e., ultrasound or x‐ray) [16]. This approach has been described previously to be aligned with data reported to EUROCAT [16].

2.5. Covariates

The following covariates were considered in the analyses: calendar year of delivery (2004–2007, 2008–2012, and 2013–2017), maternal age (< 20, 20–24, 25–29, 30–34, 35–39, and 40+), body mass index (BMI) (< 18, 18–24, 25–29, 30–34, and 35+), smoking status in first trimester (yes/no), parity (primipara/multipara), and exposure to any known teratogenic drugs (binary) during the exposure window, that is, retinoids, angiotensin‐converting enzyme inhibitor, vitamin K antagonists, valproic acid, lithium, carbamazepine, oxcarbazepine, phenytoin, phenobarbital, or methotrexate [16].

2.6. Data Analysis

Using separate study cohorts for each drug of interest, we identified associations between drug exposure and MCM among singleton livebirths for all drugs or drug classes with more than 50 livebirths exposed during the first trimester of pregnancy. A minimum cut‐off of 50 first trimester exposed livebirths was chosen to ensure sufficient sample sizes for comparative analyses. We compared exposed livebirths to unexposed livebirths in all primary analysis. In our secondary analysis, we compared exposed livebirths to livebirths of mothers treated with the same drug, that is, had filled a prescription, within 335 prior to the last day of menstruation, but not during pregnancy.

To ensure independence of outcomes, we only allowed one treated or untreated pregnancy (as appropriate) per woman in each analysis, and we excluded sibling pregnancies. The selected pregnancy for each woman was chosen at random for each specific drug analysis, and all pregnancies were eligible for re‐selection in each analysis (selection process was repeated for each drug). Using logistic regression, we estimated crude odds ratios (OR) for any MCM associated with in utero exposure to each of the study drugs and drug classes, each compared with women unexposed to the given drug during the exposure assessment window. For adjusted ORs, we constructed a multivariable model including calendar year, maternal age at delivery, BMI, smoking status, parity, and concurrent exposure to known teratogenic drugs during the first trimester, as potential confounders. The analyses of known teratogens were mutually adjusted for other known teratogens. Since this is a screening study, there was no adjustment for multiplicity.

2.7. Reporting of Results

Due to the large number of comparisons, we focused the reporting of our results on those where the outcome occurred more than twice as often in the exposed compared to the unexposed (OR ≥ 2.0). From a regulatory point (European Medicines Agency, EMA) of view, a substantially increased risk of teratogenicity (at least a doubling of background risk) is considered unlikely if no increased risk of overall MCMs is detected observed among more than 1000 individuals who were exposed to the drug during the first trimester of pregnancy [18]. This does not imply that there is no potential excess risk for those exposures with ORs less than 2.0. Importantly, as this does not rule out a teratogenic potential for drugs that only cause one or a few very specific MCMs, we also (where numbers allowed) presented results for organ‐specific MCMs separately. Consequently, we restrict our presentation of results in our tables to drug‐malformation associations with either an adjusted OR ≥ 2 or more than 1000 pregnancies exposed. For a complete overview and transparency, we provide all results (without cut‐off restrictions) and cell counts for each individual drug and cohort in Tables [Link], [Link].

Where sample size allowed, we explored organ specific MCM as defined by EUROCAT's classification of malformations. Due to the Danish interpretation of the European Union General Data Protection Regulation (GDPR), we are not allowed to extract cell counts < 5, and some outcomes are consequently not reported.

3. Results

We included a total of 743 276 singleton livebirths from 483 046 women. The selection of the study population is illustrated in Figure 2, and maternal characteristics of the cohort are tabulated in Table 1. Among 746 drugs studied, we identified 420 individual prescription drugs (ATC level 5) with ≥ 50 livebirths exposed during the first trimester of pregnancy (Table S1). The below results assessed (1) prescription drugs with potential teratogenic potential (as defined by an OR ≥ 2), and (2) prescription drugs used by more than 1000 women during pregnancy. The overall rate of MCM among 482 943 unexposed liveborn children was 3.4% and 3.6% for the total population of singleton livebirths.

FIGURE 2.

Flow diagram of cohort selection. “No prescription drug use”: Women who never redeemed any prescriptions in their life.

TABLE 1.

Cohort characteristics.

Number of livebirths	743 276 (100%)
Maternal age
Median age (IQR)	30 (25–30)
Age, < 20	9771 (1%)
Age, 20–24	81 192 (11%)
Age, 25–29	240 348 (32%)
Age, 30–34	265 554 (36%)
Age, 35–39	122 421 (16%)
Age, 40+	23 990 (3%)
Year of delivery
2004–2007	223 301 (30%)
2008–2012	267 476 (36%)
2013–2017	252 499 (34%)
Parity
Primipara	342 203 (46%)
Multipara	401 073 (54%)
Maternal pre‐pregnancy BMI
< 18 Underweight	10 941 (1%)
18–24 Normal weight	431 007 (58%)
25–29 Overweight	171 385 (23%)
30–34 Obesity class I	62 946 (8%)
35+ Obesity class II & III	38 464 (5%)
No information	28 533 (4%)
Maternal smoking status
Non‐smoker	625 580 (84%)
Smoker	97 485 (13%)
No information	20 211 (3%)
First trimester maternal drug exposure
Any prescription drug exposure	172 920 (23%)
1 prescription drug	85 727 (12%)
2 prescription drugs	42 109 (6%)
3+ prescription drugs	45 084 (6%)

3.1. Prescription Drug Exposure and Risk of Any MCM

We found ≥ 5 exposed livebirths with an MCM for 326 individual prescription drugs (Table S1), making them eligible for comparative analyses in line with regulations related to small number suppression rules to protect the anonymity of individuals. Compared to unexposed, livebirths with in utero exposure to 31 drugs during the first trimester had an increased risk of any MCM (adjusted OR [aOR] of ≥ 2.0) (Table 2). Compared to livebirths of women who discontinued treatment before pregnancy, livebirths exposed to 17 drugs during the first trimester had an increased risk of MCM (aOR ≥ 2.0). Of the 17 drugs, nine associations were also observed when comparing to the unexposed cohort (Table 2). The drugs included budesonide (oral), hydrochlorothiazide, furosemide, candesartan, dienogest, and estradiol in combination (low dose), lynesterol, diclofenac in combination with misoprostol, valproic acid, and risperidone.

TABLE 2.

Risk of major congenital malformations.

	ATC code	Drug	Cohort
			Exposed	Unexposed	Previously exposed a
			Livebirths/MCM (%)	Adjusted OR b	Adjusted OR b
1	A03BB01	Butylscopolamine	72/6 (8.3)	2.92 (1.26–6.79)	1.79 (0.61–5.25)
2	A07EA06	Budesonide	89/10 (11.2)	3.35 (1.68–6.69)	2.38 (0.82–6.92)
3	A10AB04	Insulin lispro	66/10 (15.2)	6.26 (3.16–12.4)	c
4	A10BJ02	Liraglutide	55/8 (14.5)	3.52 (1.59–7.82)	0.74 (0.20–2.70)
5	A10AE04	Insulin glargine	460/45 (9.8)	2.71 (1.98–3.72)	0.93 (0.39–2.24)
6	A10AB01	Insulin (human)	601/49 (8.2)	2.59 (1.91–3.51)	0.61 (0.29–1.31)
7	A10AC01	Insulin (human)	1069/86 (8.0)	2.53 (2.01–3.17)	0.81 (0.44–1.50)
8	A10AB05	Insulin aspart	1495/125 (8.4)	2.43 (2.01–2.93)	1.34 (0.52–3.46)
9	A10AD05	Insulin aspart	237/19 (8.0)	2.24 (1.40–3.59)	c
10	A10AE05	Insulin detemir	473/35 (7.4)	2.12 (1.48–3.02)	1.09 (0.46–2.62)
11	B01AC04	Clopidogrel	68/8 (11.8)	3.90 (1.86–8.20)	c
12	B03AA01	Ferrous glycine sulfate	52/6 (11.5)	4.17 (1.77–9.83)	c
13	C03DA01	Spironolactone	64/7 (10.9)	2.94 (1.26–6.85)	c
14	C03EA01	Hydrochlorothiazide and potassium‐sparing agents	59/5 (8.5)	2.88 (1.15–7.26)	2.34 (0.57–9.54)
15	C03CA01	Furosemide	256/24 (9.4)	2.61 (1.68–4.05)	2.42 (1.27–4.60)
16	C09CA06	Candesartan	101/9 (8.9)	2.88 (1.45–5.72)	5.29 (1.48–18.9)
17	C09BA02	Enalapril and diuretics	54/6 (11.1)	2.45 (1.02–5.92)	c
18	C09DA01	Losartan and diuretics	59/5 (8.5)	2.39 (0.95–5.99)	c
19	D01AE14	Ciclopirox	127/6 (4.7)	1.66 (0.73–3.77)	2.55 (0.80–8.13)
20	G03AB08	Dienogest and estradiol	57/5 (8.8)	2.71 (1.08–6.82)	2.24 (0.78–6.40)
21	G03DC03	Lynestrenol	79/5 (6.3)	2.14 (0.86–5.30)	2.14 (0.67–6.85)
22	H03BB01	Carbimazole	55/5 (9.1)	3.21 (1.28–8.09)	c
23	H04AA01	Glucagon	756/57 (7.5)	2.23 (1.69–2.94)	0.81 (0.45–1.45)
24	J01DC02	Cefuroxime	81/5 (6.2)	1.89 (0.76–4.68)	4.50 (0.76–26.6)
25	M01AB55	Diclofenac, combinations	91/7 (7.7)	2.45 (1.13–5.32)	4.44 (1.62–12.2)
26	M01AE03	Ketoprofen	100/6 (6.0)	1.99 (0.87–4.56)	2.96 (0.96–9.15)
27	M01AC01	Piroxicam	101/5 (5.0)	1.77 (0.72–4.36)	2.12 (0.72–6.25)
28	N03AG01	Valproic acid	235/18 (7.7)	2.07 (1.22–3.50)	3.27 (1.10–9.75)
29	N03AX16	Pregabalin	296/22 (7.4)	2.01 (1.29–3.14)	1.77 (0.90–3.46)
30	N05BA09	Clobazam	84/8 (9.5)	3.15 (1.50–6.60)	c
31	N05CD02	Nitrazepam	113/9 (8.0)	2.26 (1.10–4.66)	1.13 (0.38–3.39)
32	N05AX08	Risperidone	157/12 (7.6)	2.22 (1.23–4.01)	2.55 (1.07–6.10)
33	N05AF03	Chlorprothixene	368/26 (7.1)	2.05 (1.36–3.09)	1.75 (1.03–2.98)
34	N05BB01	Hydroxyzine	98/5 (5.1)	1.54 (0.63–3.80)	2.13 (0.62–7.32)
35	N06AX03	Mianserin	225/13 (5.8)	1.75 (1.00–3.08)	2.10 (1.02–4.32)
36	N07BB01	Disulfiram	90/8 (8.9)	2.71 (1.24–5.91)	1.27 (0.45–3.58)
37	P01BA02	Hydroxychloroquine	261/12 (4.6)	1.23 (0.67–2.25)	2.30 (0.75–7.08)
38	R01AD11	Triamcinolone	230/14 (6.1)	2.08 (1.21–3.58)	1.61 (0.75–3.44)
39	S01AD03	Aciclovir	143/7 (4.9)	1.37 (0.60–3.10)	2.48 (0.73–8.39)

Note: Drugs with an OR ≥ 2 (regardless of estimated precision) compared against either unexposed live births and/or live births to mothers exposed prior to pregnancy.

Abbreviations: ATC, anatomical therapeutic chemical; MCM, major congenital malformations; OR, odds ratio.

^a Maternal exposure within 335 days prior to 30 days before last menstrual period only.

^b Adjusted for body mass index, smoking, calendar year, teratogenic drug exposure, parity, and maternal age.

^c Cannot be estimated due to low numbers.

Twenty‐one observed associations confirmed previously identified teratogenic drugs, including insulins (likely secondary to diabetes), ace‐inhibitors and angiotensin II receptor blockers, valproic acid, and pregabalin. Other known teratogenic drugs were not captured in the analysis due to a low number of exposed pregnancies.

For 18 drugs, our data suggest potential teratogenic effects where such are not well established: butylscopolamine, budesonide (oral), clopidogrel, ferrous glycine sulfate, spironolactone, hydrochlorothiazide, furosemide, dionegest and estradiol in combination (low dose), lynesterol, glucagon, cefuroxime, clobazam, nitrazepam, mianserin, disulfiram, triamcinolone (nasal inhalation formulation), and acyclovir (dermal formulation). For some of these associations, the number of exposed cases was very low.

3.2. Exposure to Drugs Prescribed to More Than 1000 Livebirths During First Trimester and Risk of MCM

We identified 115 drugs prescribed more than 1000 times either before or during pregnancy (Table S3) making for a more precise risk estimate. Thirty‐three drugs had an aOR above 1 (point estimate between 1.10 and 2.53) with a lower‐bound confidence interval above one when compared against unexposed. When comparing against previously exposed, all estimates were attenuated, and only pivmecillinam, prednisolone, and lansoprazole maintained a CI with a lower‐bound above 1.

3.3. Organ Specific MCM

Drug‐organ specific MCM associations are shown in Table S4. The most common organ specific MCM were cardiovascular, followed by urological, nervous system, limb, gastrointestinal, and cleft palate malformations. We identified 84 drugs with an aOR ≥ 2.0 (regardless of precision) for any organ specific MCM (at least 5 organ specific MCM among exposed), when compared to unexposed live births. When comparing with live births of mothers who discontinued treatment before conception, 54 drugs had an aOR ≥ 2.0, and 14 drugs had an aOR ≥ 2.0 compared to both comparator groups. Of drugs prescribed more than 1000 times but with no overall MCM, 11 drugs had an aOR ≥ 2.0 for organ specific MCM compared to either unexposed live births or live births of women exposed prior to pregnancy.

4. Discussion

4.1. Principal Findings

4.1.1. Drugs With an Increased Risk of MCM

In this nationwide register‐based crude screening study of all singleton livebirths in Denmark between 2004 and 2017, we identified 420 individual drugs with ≥ 50 livebirths exposed during the first trimester of pregnancy. We found 326 potential drug‐malformation candidates for any MCM, of which 31 had aOR ≥ 2.0 when comparing to unexposed livebirths. Among 115 prescription drugs with ≥ 1000 livebirths exposed in the first trimester, we, importantly, found most drugs had no or minor increased risk (as defined by an OR < 2.0) of MCM following first trimester in utero exposure. Among these, 11 drugs had increased risk for organ‐specific MCM. Only insulins had an aOR ≥ 2.0 for MCM when comparing to unexposed livebirths. All individual drug associations for overall MCM were substantially attenuated when comparing livebirths to mothers who received treatment prior to conception with the identical drug but not during pregnancy. Some (weak) associations persisted for pivmecillinam, prednisolone, and lansoprazole, suggesting that the associations found when comparing against unexposed livebirths were confounded by factors associated with the underlying disease. These pivmecillinam and lansoprazole are well documented not to be associated with an increased risk of congenital malformations when studied in a disease‐specific context, while data remain ambiguous for prednisolone [19, 20, 21]. There was no apparent consistent effect on the precision of the estimates.

4.1.2. Drugs Prescribed to More Than 1000 Livebirths During First Trimester

We identified 112 drugs that complies with regulatory (European Medicines Agency) guidelines to substantiate, with a high degree of precision, that an association between first trimester exposure and overall MCM is unlikely. This includes drugs prescribed to more than 1000 women during pregnancy, such as most proton‐pump inhibitors, inhaled anti‐asthmatic drugs, oral antihistamines, oral anticontraceptive drugs, most penicillins and macrolides, non‐steroid anti‐inflammatory drugs (NSAIDs), selective serotonergic reuptake inhibitors (SSRIs), metoclopramide and metformin.

4.2. Strengths and Limitations

We demonstrated the feasibility of a nationwide screening model that allows for the detection of associations of MCM following first trimester exposure to all marketed prescription drugs in Denmark. The associations were qualified through several steps including the use of different comparison groups and multiple adjustments in a series of individual cohort studies. Our model identified well‐established teratogenic drugs including valproic acid, carbimazole, risperidone, and candesartan. These serve as external positive controls, and our similar findings represent a reassuring external validation of the model.

As in all pharmacoepidemiologic register‐based studies we know that the prescriptions were filled, but do not know if women took the drugs. We have no information on over‐the‐counter medication use or use of recreational substances. Reliable data on alcohol use in pregnancy is not available. Also, high mortality from MCM would likely lead to very early pregnancy loss and therefore be underreported. A principal limitation of our study is the exclusion of non‐livebirth pregnancy outcomes due to inconsistency and low validity of this information but in a quantitative bias analysis, restricting to livebirths resulted in negligible differences in estimates the association between drug exposures and birth defects [22]. In this hypothesis generating approach, we did not address the use of more than one drug during one pregnancy. We acknowledge a substantial risk of false positive associations inherent to studying more than 700 potential associations and we encourage cautious interpretation, especially for inferential analyses based on a low number of outcomes as reflected in imprecision of the accompanying estimates. Dichotomizing data according to traditional hypothesis testing (i.e., using p‐values) [23, 24, 25, 26] and associated corrections for multiple testing are not useful in these types of screening studies [27]. We emphasize that the associated upper and lower confidence limits represent compatibility with the data and the possible effect sizes associated with MCM following first trimester exposure [23, 24, 25].

5. Results in Context

We replicated findings of known teratogenic drug exposures (e.g., insulin, risperidone, valproic acid, angiotensin converting enzyme inhibitors, and carbimazole) as well as potential unidentified drugs with teratogenic potential (e.g., clopidogrel, furosemide, mianserin, and insulin detemir) [28, 29, 30]. We confirmed multiple drugs with no apparent increased risk of MCM after exposure during the first trimester (e.g., inhaled corticosteroids, many antibiotics, and many antihistaminergic drugs) [19, 31, 32].

6. Clinical Implications

Our findings provide substantial evidence for everyday decision support and shared decision making. Specifically, this pertains to 112 null findings that all align with regulatory (European Medicines Agency) guidelines for stating that an association between first trimester exposure and overall MCM is unlikely for these drugs.

7. Research Implications

The identified associations suggesting potential teratogenic effects requires confirmation and calls for additional studies. This could be implemented for each drug in pharmacoepidemiologic studies addressing disease or disorder specific confounders and covariates and patient characteristics while (preferably) using an active comparator group. Future analysis will include more recent data. Ongoing work aims to establish a Core Data Model which would allow the developed approach herein to be applied to the restructured Danish National Registries. We emphasize that point estimates alone should not form the basis for treatment decisions, but include considerations on precision, absolute risks and risks related how untreated medical conditions can affect both mother and fetus should be considered in its totality. Another general statistical approach for post‐marketing data mining to study adverse effects of drugs, Treescan was developed by the Food and Drug Administration and Centers for Disease Control and prevention [33]. Huybrechts et al. [34] recently successfully validated the applicability of this hierarchical datamining approach to the use of opioids and valproate in pregnancy specifically for associations to neonatal withdrawal and specific MCM, respectively. This model has not been applied or validated across a broader spectrum of drugs.

8. Conclusion

In conclusion, we identified 112 drugs prescribed to more than 1000 women during first trimester with no overall increased risk of congenital malformations among live‐born children. We identified 18 unestablished potential associations including butylscopolamine, budesonide (oral), clopidogrel, furosemide, spironolactone, cefuroxime and mianserin. These findings should be substantiated in studies allowing for disease specific confounder‐control with an active comparator design.

8.1. Plain Language Summary

Prescription drug use during pregnancy: Advising pregnant women or those planning pregnancy about prescription medications is challenging due to potential risks to the unborn child. This study focused on the risk of serious birth defects after first‐trimester use of prescription medication in Denmark from 2004 to 2017, using various data sources to compare the risk of serious birth defects in newborns of mothers using prescription medication during pregnancy. Key findings: Out of 326 medications, 31 were linked to an increased risk of birth defects. Compared to women who stopped medication before pregnancy, 17 medications were linked to higher risks. Only insulins had a significant risk among 115 drugs prescribed to > 1000 women. More than 100 medications showed no increased risk. Limitations: This study was a screening study of all prescription drugs and did not focus on disease‐specific risk factors; however, it did include common risk factors for birth defects. Conclusions: Over 100 medications were not linked to serious birth defects, providing reassurance for pregnant women. Some harmful medications were confirmed, and new ones identified for further study.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Risk of major congenital malformations in exposed livebirths compared to unexposed livebirths and livebirths to mothers exposed prior to pregnancy. All individual prescriptions drugs (and ATC level 3 and 4 categories) with at least 50 infants with first trimester exposure.

Table S2: Risk of major congenital malformations. Drugs with at least 1000 first trimester exposed livebirths OR livebirths to mothers exposed prior to pregnancy. Risk for exposed livebirths compared against unexposed livebirths and livebirths to mothers treated prior to pregnancy.

Table S3: Risk of major congenital malformations. Drugs with at least 1000 first trimester exposed livebirths compared to unexposed livebirths and livebirths among mothers exposed prior to pregnancy.

Table S4: Organ specific major congenital malformations (at least 5 observed among exposed livebirths) with aOR ≥ 2.0 (regardless of estimate precision) for comparison against either unexposed livebirths or livebirths whose mother was treated prior to pregnancy.

Broe A., Pottegård A., Munk‐Olsen T., et al., “Prescription Drugs in Pregnancy and Congenital Malformations: A Population‐Based Safety Screening Study,” Pharmacoepidemiology and Drug Safety 34, no. 9 (2025): e70211, 10.1002/pds.70211.

Data Availability Statement

The authors have nothing to report.