Study population

The study population was derived from the NBDPS; a detailed description of the study design has been published elsewhere. Briefly, the NBDPS is an ongoing case-control surveillance study that was designed to identify infants with major birth defects and to evaluate genetic and environmental risk factors. Case infants for the study are identified from the state birth defect surveillance system of 10 states (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, New York, Texas, North Carolina, and Utah). Cases include live births, stillbirths of ≥20 weeks’ gestation or >500 g, or elective terminations with any of >30 eligible birth defects groups that are diagnosed in the first year of life. Each case is reviewed and confirmed by a clinical geneticist. Control subjects are liveborn infants with no major birth defects who are selected from the same base population as the cases. Control subjects are selected randomly either from birth certificates or from birth hospitals with the use of a stratified random sample. Our study included women with expected due dates between October 1, 1997, through December 2004. The NBDPS is a population-based study. All cases that meet study criteria and who reside in the study area are study eligible. All randomly selected control subjects are also study eligible. Participation rates represent the proportion of subjects who are study eligible who agreed to participate in the study. Of all control subjects who were asked to participate in the study, 69% agreed; of all cases who were asked to participate in the study, 72% agreed. Reasons for nonparticipation included women who were contacted and declined to participate and women who could not be located before they aged out of the study (>24 months from estimated due date). All study sites obtained institutional review board approval; all study participants provided informed consent.

Exposure

Mothers are contacted and interviewed by way of a computer-assisted telephone interview that lasts approximately 1 hour. The interview includes detailed questions about exposures that occurred 3 months before conception through the end of pregnancy. A pregnancy calendar that is built into the computer-assisted telephone interview allows the mothers to respond about the timing of exposures in the way that she best recalls: by date, month of pregnancy, or trimester. Interviews are to be completed no earlier than 6 weeks and no later than 24 months after the estimated due date. Mothers are asked whether they had any of a list of specific illnesses or conditions (eg, cold, flu, injuries); if so, what medications, if any, were taken. The mothers were also asked about use of Advil (Pfizer Inc, New York, NY), Motrin (McNeil-PPC, Inc, Philadelphia, PA), Nuprin (Bristol-Meyers Squibb, Princeton, NJ), ibuprofen, aspirin, and Aleve (Bayer Healthcare LLC, Morristown, NJ). For each medication that was reported, mothers were asked when they took it and how often. If a woman reported that she had taken a medication and reported a start date that was in the first trimester , but could not characterize frequency of use in terms of daily or days per week or month, her use was considered to be “as needed.” All medications that were reported in the interview were coded according to the Slone Epidemiology Center Drug Dictionary (Slone Epidemiology Center, Boston, MA), which links product name to all components.

Because most structural malformations develop within the first trimester, we focused on NSAID use during this time period, which was defined as the estimated day of conception through 3 months afterward. NSAIDs included medications with the following ingredients: carboxyl acetic acid, carboxyl anthranilic acid, carboxyl propionic acid, carboxyl salicylate, enolic pyrazole, enolic oxicam, enolic pyrazolone, enolic quinazolinone, and other miscellaneous components. NSAIDS were further grouped according to component into aspirin, ibuprofen, or naproxen; there were not enough women who used other types of NSAIDs to examine them separately. Women who reported taking multiple types of NSAIDs were included in all applicable exposure groups. Because the month of consumption could not be categorized with certainty for medications that were reported “as needed,” we considered such use separately. Thus, for each medication, women were divided into 3 mutually exclusive groups: exposed in the first trimester, as needed, and unexposed at any time during pregnancy (P1-P9). In addition, we examined multiple and single component NSAID use, as well as route of exposure (oral or topical).

Outcome

Clinical geneticists at each site reviewed clinical data on potential cases with the use of standardized case definitions to ensure their study eligibility. Cases with known cause (single-gene disorders or chromosomal abnormalities) were excluded. Cases were further classified according to the presence or absence of other structural defects. Cardiac defect cases were also classified as simple if they were discrete anatomically.

Over 30 structural malformation groups are included in the NBDPS. We restricted our analysis to defects with ≥100 cases, after excluding those cases with a family history of the same type of malformation, to have sufficient power to examine potential risk factors. Noncardiac case groups that were eligible for analysis included anencephaly and craniorachischisis, spina bifida, encephalocele, hydrocephaly, anophthalmia/microphthalmia, cataracts, anotia/microtia, cleft palate with or without cleft lip, cleft palate alone, esophageal atresia, intestinal atresia/stenosis (only jejunal, ileal, and multiple small-intestinal atresias), anorectal atresia/stenosis, hypospadias (second/third degree), longitudinal limb deficiency, transverse limb deficiency, craniosynostosis, diaphragmatic hernia, omphalocele, gastroschisis, and amniotic band syndrome/limb body wall complex. For cardiovascular defects, cases were restricted to those classified as simple (ie, well-defined, pure conditions), without accompanying noncardiac defects and included dextro-rotated transposition of the great arteries, tetralogy of Fallot, ventricular septal defect (perimembraneous), ventricular septal defect (muscular-only collected for the first year of the study), pulmonary valve stenosis, aortic stenosis, hypoplastic left heart syndrome, coarctation of the aorta, and total anomalous pulmonary venous return.

Inclusion/exclusion criteria

We excluded mothers with preexisting diabetes mellitus for both cases (n = 709) and control subjects (n = 130) because of the known association of diabetes mellitus with various birth defects. We also excluded women who had missing information about their exposure to NSAIDS or were exposed only during trimesters 2 or 3 (cases, 799 women; control subjects, 282), which yielded a final study size of 14,915 cases and 5546 control subjects.

Statistical analysis

We first examined the distribution of characteristics between various exposure groups among control subjects, and also the distribution of characteristics between cases and control subjects. Descriptive characteristics included interview site, maternal race/ethnicity, maternal age at delivery, maternal body mass index, maternal highest education level achieved, folic acid supplementation during the period 1 month before conception through the first month of the first trimester, maternal smoking (from 1 month before conception through the end of the first trimester), sex of infant, binge drinking, gestational diabetes mellitus, fever (from 3 months before pregnancy through the end of first trimester), time between birth and interview, multiple births, and parity. Fever was ascertained through questions regarding the occurrence of fever that was associated with certain illnesses (such as respiratory tract infections and urinary tract infections) and through explicitly asked questions about fever independent of any other illnesses. Confounders that were selected were those that were associated with exposure among the control subjects or associated with the outcome (study site, race/ethnicity, folic acid, smoking, binge drinking, and fever in the first trimester). We used logistic regression to estimate crude and adjusted odds ratios (ORs) and their 95% confidence intervals (CIs).