Association Between Proton Pump Inhibitor Use During Early Pregnancy and Risk of Congenital Malformations

Key Points Question Is proton pump inhibitor (PPI) use during pregnancy associated with an increased risk of congenital malformations? Findings In this cohort study including 2 696 216 pregnancies in South Korea from 2011 to 2019, PPI use during the first trimester was not associated with a substantial increase in the risk of major congenital malformations, congenital heart defects, cleft palate, hydrocephalus, and hypospadias. Findings from the sibling-controlled analyses also revealed that PPIs were unlikely to be a major teratogen. Meaning These findings may help guide clinicians and patients in decision-making about PPI use in early pregnancy.

To increase the specificity, we defined the presence of congenital malformations based on the following algorithms.

Major congenital malformations:
• Presence of any of the malformations stated below 12 System-specific malformations (e.g. congenital heart defects): • ≥2 visits with any of the codes for organ-specific malformation as a primary diagnosis within one year after delivery (regardless of inpatient or outpatient visits) • ≥1 visit with a code for organ-specific malformation as a primary diagnosis and malformationspecific surgery/procedure code within one year after delivery • ≥1 visit with a code for organ-specific malformation as a primary diagnosis and death within one year after delivery Subgroup of system-specific malformations (e.g. cleft palate, hydrocephalus, hypospadias): • ≥2 visits with any of the codes for the subgroup of organ-specific malformation as a primary diagnosis within one year after delivery (regardless of inpatient or outpatient visits) • ≥1 visit with a code for the subgroup of organ-specific malformation as a primary diagnosis and malformation-specific surgery/procedure code within one year after delivery • ≥1 visit with a code for the subgroup of organ-specific malformation as a primary diagnosis and death within one year after delivery To further evaluate the reliability of our outcome definition, we tested whether we could reproduce the known associations between pre-existing diabetes and the congenital malformations in our data. 1 Preexisting diabetes was defined as having 2 or more diagnostic codes between 180 days prior to the start of pregnancy and the first trimester. As a result, we observed significant associations for both major congenital malformations (RR 2.00, 95% CI 1.89-2.12) and congenital heart defects (2.60, 2.42-2.79).
We also conducted additional analyses that restricted the outcome definition to inpatient diagnoses only and the estimates were consistent with our main result (major congenital malformation: adjusted RR 1.07, 95% CI 0.99-1.15; congenital heart defects: 1.03, 0.93-1.14).

eAppendix 2. Additional Details on the Sibling Analysis
To account for potential confounding from family-related factors, we additionally performed siblingcontrolled analyses. In the sibling analysis, shared familial/genetic factors within the family could be adjusted by comparing the risk of outcome among the infants born to the same mother. In our study, we included women with at least two pregnancy episodes during the study period. Among those pregnancies, only the siblings who were discordant for both exposure and outcome contributed to the estimates. Because the estimation depends on the discordant sibling pairs, multifetal pregnancies did not contribute to the estimate as they share the same exposure status. Using the logistic regression models stratified on the mother's unique identifier, odds ratio (OR) with 95% confidence intervals (CIs) were estimated adjusting for all covariates considered in the main analysis. The results are presented in Figure 3 and because there were zero cases observed among the exposed male siblings, the risk of hypospadias could not be estimated.
Additional analyses were done to test the assumptions of the sibling analysis: First, to confirm whether the results from sibling populations are generalizable to the full population, we reran the analysis among the sibling populations while not using the stratified term. Overall, we found comparable results between the sibling populations and the full population in our main analysis (eTable 8).
Second, the sibling analysis assumes no carryover effect, that is the outcome of the first sibling should not influence the exposure status of the second sibling 1 . To test the presence of carryover effect, we analyzed the association between the congenital malformation status of the first sibling and PPI exposure during the second pregnancy. The OR of PPI exposure during the second pregnancy was estimated according to the malformation status of the first sibling, while controlling for PPI exposure during the first pregnancy. As a result, we obtained odds ratio approximate to 1 for all our outcomes, indicating that the carryover effect is most likely absent (eTable 9).

eAppendix 3. Potential Consequences of Including Only Live Births
Our study cohort included pregnancies that resulted in live births and did not include pregnancies that ended in stillbirth or abortions. This restriction may introduce selection bias if the probability of live birth differs between PPI-exposed and unexposed pregnancies. For instance, if the probability of live birth is lower in PPI-exposed pregnancies than in unexposed pregnancies owing to a higher rate of pregnancy terminations due to severe malformations, then the risk estimates may be biased toward the null. Therefore, we quantified the potential effects of missing non-live births.
Considering the different probability of live births between PPI-exposed and unexposed pregnancies, the corrected relative risks were estimated as below. This method has been widely used in previous studies to quantify the potential effect of restriction to live births 1,2 : Corrected relative risk (RR)=Observed RR*(S10*S01/S11*S00) S10 refers to the probability of live births in PPI-unexposed pregnancies with malformation.
S01 refers to the probability of live births in PPI-exposed pregnancies without malformation. S11 refers to the probability of live births in PPI-exposed pregnancies with malformation. S00 refers to the probability of live births in PPI-unexposed pregnancies without malformation.
Based on estimates from the literature, the live birth probability among PPI-unexposed pregnancies without malformation (S00) was defined as 80%. 3 We then assumed the probability of live birth among unexposed pregnancies with malformations (S10) as a range of 55% to 80%, based on a previous study. 4 Lastly, we evaluated the potential effect of lower frequency of live births, ranging from 10% to 20%, in PPI exposed pregnancies (eTable 10).