The BWIS population consisted of infants born to residents of Maryland, the District of Columbia, and 6 adjacent counties of northern Virginia from April 1981 through December 1989. The methods of this study have been previously described in detail.^10,11 All data used in our study were deidentified and analyses were performed with an exemption from the Institutional Review Board of the Centers for Disease Control and Prevention.

We found that use of any analgesic during the periconceptional period was common among pregnant women enrolled in BWIS, with the most commonly used analgesic class being medications containing acetaminophen. Analgesic use did not differ by the presence of fever or flu symptoms. Consistent with prior knowledge of congenital cardiac defects,¹⁴ we found that family history of cardiac malformation was associated with increased prevalence of CVM in the offspring. Although analgesic use in the periconceptional period was not associated with CVM in the aggregate, we found associations of specific CVM phenotypes with maternal periconceptional use of acetaminophen and NSAID.

An association of CVM with periconceptional NSAID or acetaminophen use is consistent with the hypothesis that COX inhibition during fetal heart development might increase the risk of CVM in the infant. However, if COX inhibition were the underlying cause, one would expect that fetal exposure to irreversible inhibition of COX isoforms by salicylate during the critical period would also result in CVM. We did not observe an association between salicylate use and CVM.

Prior studies examining the association between exposure to COX inhibitors during the critical period of heart development and CVM have had mixed findings. In animal studies, use of high-dose aspirin and NSAIDs has been associated with cardiac septal defects. ^15,16 In human beings, data from the Collaborative Perinatal Project, a prospective cohort study, showed no association of aspirin use early in pregnancy with CVM in the aggregate.¹⁷ Similarly, findings from a case-control study by Werler et al¹⁸ showed no association of aspirin use with CVM in the aggregate (relative risk, 0.9; 95% CI, 0.8–1.1). However, a slight association between maternal NSAID use during gestational weeks 5–8 and atrial septal defects was seen in a prospective cohort study from Norway, although the findings were not statistically significant (adjOR, 1.6; 95% CI, 0.7–3.9).¹⁹ Data based on the Swedish Birth Registry suggested an association between NSAID use and CVM in the aggregate (OR, 1.86; 95% CI, 1.32–2.62).²⁰ A Canadian Registry–based study also found an association between maternal NSAID use in the first trimester and cardiac septal defects (OR, 3.34; 95% CI, 1.87–5.98).²¹

Prospective data from the Danish National Birth Cohort and case-control data from the National Birth Defects Prevention Study (NBDPS) showed no increased risk of CVM with acetaminophen use.^22,23 In addition, no significant associations were noted between muscular VSD and maternal use of either NSAIDs or acetaminophen, adjusting for maternal fever, in NBDPS data.²⁴ However, more recent reports from this study have shown associations of maternal use of opioid analgesics with all CVM in the aggregate as well as with specific cardiac defects (ie, AVSD without Down syndrome, tetralogy of Fallot, conoventricular septal defect, hypoplastic left heart syndrome, pulmonary valve stenosis, and atrial septal defect + VSD and pulmonary valve stenosis + VSD associations), and of maternal use of naproxen with isolated pulmonary valve stenosis among full-term infants. ^25,26 A recent study from The Norwegian Mother and Child Cohort Study found no association between maternal codeine use in the first trimester and congenital anomalies in the offspring.²⁷

Previous analyses of BWIS data focusing on possible risk factors for each cardiac phenotype found associations of maternal periconceptional use of ibuprofen (1 type of NSAID) with dTGA with intact ventricular septum and with AVSD among infants with Down syndrome, and use of aspirin with interrupted aortic arch.^10,28 We did not include interrupted aortic arch in our study due to its strong association with DiGeorge syndrome. Our analysis, which excluded infants with chromosomal anomalies and any noncardiac birth defects, also showed an association between dTGA with intact ventricular septum and maternal NSAID use.

Among the cardiac phenotypes common to the recent report on opioid exposure from NBDPS and ours (ie, AVSD without Down syndrome, tetralogy of Fallot, hypoplastic left heart syndrome, and pulmonary valve stenosis), we found no associations with opioid analgesics. Possible reasons for the difference in findings between the 2 studies include a smaller sample size for each specific cardiac phenotype in our study; possible differences in composition between the opioid analgesics used during the different time periods (ie, 1981 through 1989 vs 1997 through 2005), the amount of opioid used, concomitant use of other medications, or the indication for use; and possible sampling variation.

Our findings need to be viewed in light of the limitations and strengths of the study. Limitations include potential maternal recall bias due to self-reported exposure status. However, >90% of the home interviews for both cases and controls were completed within 12 months of the infant’s birth and the interview questions related drug use to specific medical indications.^10,29 A low likelihood of recall bias is also suggested by a previous analysis of reported drug use within the BWIS data, which showed no significant differences between the number of drugs reported by the time elapsed between delivery and interview. ¹³ As the time period of exposure spans up to 3 months before the last menstrual period, there may be a bias towards the null, as the half-life of most analgesics is <30 hours.³⁰ Another limitation pertains to the lack of information regarding the maternal dose of analgesic used, which prevented us from examining dose-response for any of the pharmacologic classes of analgesics. There were also a small number of women using isolated opioid medications during the preconceptional period in our study, which prohibited us from assessing associations of many CVM types with maternal opioid use. We did not have information on some of the newer analgesic classes, such as COX-2 inhibitors, which were not widely used during the period of the BWIS. In addition, we included mothers who used nonanalgesic medications and could not control for possible differential exposure to these drugs between case and control infants. We were not able to assess any potential joint effects of these preparations because of the small number of mothers who reported taking individual multicomponent preparations. Confounding by indication is another potential limitation.

There were also some nonspecific categories of medication use reported in the BWIS data, such as miscellaneous cold medicines, and some reports of medication use for which the drug name was not known; these were not included in our exposure definition but might have contained analgesics and could have resulted in exposure misclassification. To assess the impact of maternal illness, we performed analyses controlling for the presence of maternal fever or flu symptoms as proxies for hyperthermia and influenza. However, some maternal infections might have been subclinical or underreported resulting in residual confounding. Also, we made no statistical corrections for multiple comparisons, which may have resulted in our finding of 2 significant associations due to chance alone. With the number of statistical comparisons made, approximately 4 significant findings would be expected due to chance and not to truly significant associations. Finally, the BWIS data are >20 years old, and the pattern of analgesic use may have changed over that time. However, use of analgesics is still common and indications for use have not changed significantly during the intervening period.

Strengths of our study include the population-based design of BWIS, which can minimize the potential for selection bias. Also, cases were identified from multiple sources, which maximized ascertainment. Other strengths are the classification of CVM in this study that took embryonic origin into account and was reviewed by pediatric cardiologists, and the verification of CVM diagnoses by echocardiogram, catheterization, or surgery for all case infants, which allowed evaluation of possible associations of analgesic use with specific phenotypes. In addition, we classified all analgesics by their pharmacologic mechanism, which allowed evaluation of biologically plausible associations between maternal analgesic use and CVM. Finally, we examined maternal analgesic use during the periconceptional period of organogenesis when potential teratogens would be expected to affect cardiac development.

Our findings provide reassurance that maternal analgesic use during pregnancy does not result in major teratogenic effects relative to congenital heart defects. We did find suggestion of a possible association of maternal use of NSAIDs and acetaminophen during the periconceptional period with dTGA with intact ventricular septum and tetralogy of Fallot in the offspring, respectively. Further studies are warranted to replicate these findings. Future studies regarding the potential teratogenicity of analgesics also should include information on dose and the specific timing of use; evaluate the use of newer analgesic classes such as COX-2 inhibitors; and attempt to limit exposure to the period of cardiac development while taking into account the half-life of the drugs, their excretion, and their indication for use.