Orofacial clefts are one of the most common birth defects in humans, with a world birth prevalence of 1.7 per 1000 live births (Mossey et al., 2009). Orofacial clefts occur when the fusion of the lip and/or palate, which takes place during the first-trimester of pregnancy, is disrupted (Dixon et al., 2011). Corticosteroids are well-established as an experimental teratogen in animal models, causing cleft palate in mice (Fraser and Fainstat, 1951; Walker and Fraser, 1957). Several epidemiological studies have reported an association between corticosteroid use in early pregnancy in humans and delivering an infant with an orofacial cleft (Czeizel and Rockenbauer, 1997; Rodríguez-Pinilla and Luisa Martínez-Frías, 1998; Carmichael and Shaw, 1999; Edwards et al., 2003; Pradat et al., 2003; Carmichael et al., 2007), although others have not (Kallen et al., 1999; Källén, 2003; Hviid and Mølgaard-Nielsen, 2011).

The anti-inflammatory and immune modulating functions of corticosteroids are effective in the treatment of conditions such as asthma, allergic reactions, eczema, psoriasis, rheumatoid arthritis, and inflammatory bowel disease. These conditions are common and often affect women of reproductive age; however, the safety of corticosteroid medication during pregnancy is uncertain.

We previously reported that maternal corticosteroid use was associated with increased risk of cleft lip with or without palate (CLP) (odds ratio [OR], 1.7; 95% confidence interval [CI], 1.1–2.6) but not cleft palate only (CPO) (OR, 0.5; 95%CI, 0.2–1.3), using data from the National Birth Defects Prevention Study (NBDPS) investigating deliveries from October 1997 through December 2002, including mothers of 1141 infants with CLP, 628 infants with CPO and 4143 controls (Carmichael et al., 2007). Since then, the study population has more than doubled in size, allowing the largest study of corticosteroids and clefts to date. Given continued uncertainty about the association between orofacial clefts and corticosteroid medications and the tentative findings from our earlier analyses, our objective here was to assess the association using larger and more recent NBDPS data.

We used data from the NBDPS, a population-based, multi-center case–control study of birth defects. Information on deliveries taking place from October 1997 through December 2009 was collected from the 10 NBDPS study centers (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, New York, North Carolina, Texas, and Utah), although not all study sites contributed for all the study years. The study was approved by institutional review boards of the participating centers and the Centers for Disease Control and Prevention. More details on study methods and its surveillance systems can be found elsewhere (Yoon et al., 2001; Rasmussen et al., 2003).

Infants or fetuses with CLP or CPO were considered cases and analyzed separately. Case status was ascertained either through clinical or surgical records or autopsy reports. Medical records for all cases were assessed by a clinical geneticist who ensured that they fulfilled the eligibility criteria. Cases were ineligible if their clefts were believed to result from another defect (e.g., holoprosencephaly) or had a recognized or strongly suspected single-gene disorder or chromosomal abnormality. Cases were considered isolated if there were no accompanying major unrelated birth defects or as nonisolated if more than one additional major unrelated defect was present. Controls (live born infants, without birth defects) were randomly selected from hospital birth records or birth certificates at each study center.

Mothers were interviewed 6 weeks to 24 months after estimated date of delivery, using computer-assisted telephone interviews in English or Spanish; median time between estimated date of delivery and interview was 9.0 months for case mothers (interquartile range 8.0 months) and 8.0 months for controls (interquartile range 7.0 months). Overall participation from 1997 to 2009 was 72% for eligible mothers of infants with clefts and 65% for control mothers (participation in the two time periods declined from 76% to 67% for eligible mothers of infants with clefts and 68% to 61% for control mothers, with an overall decline from 70% to 63%).

In the questionnaires the mothers were asked whether they had specific medical conditions before or during pregnancy and then what medications they used to treat them. Mothers were also asked to list any other medication they had used that was not captured in response to the specific questions; indication was not reported for responses to this question. Mothers were asked for duration and frequency of use for each medication used from 12 weeks before conception to delivery. Medications were coded according to the Slone Epidemiology Center Drug Dictionary. We focused on periconceptional corticosteroid use by any administration route and component (systemic, nasal/inhaled and topical), defined as use that occurred between 4 weeks before through 12 weeks after conception.

We investigated the association between any corticosteroid use during the periconceptional period compared with no use. We also explored whether there was an association with specific timing of exposure, mode of administration, or corticosteroid component. Logistic regression models in SAS software were used to estimate ORs and their corresponding 95% CIs. ORs were only calculated if there were two or more exposed cases and two or more exposed controls. We also examined associations after adjustment for several covariates (maternal race-ethnicity, education, intake of folic acid-containing supplements, smoking, and study center) and after exclusion of nonisolated cases. We present results for deliveries from January 2003 through December 2009, and for pooled data for deliveries from October 1997 through December 2009.

From 2003 to 2009, the NBDPS enrolled mothers of 1577 children with CLP, 795 children with CPO, and 5922 control children. Demographic characteristics are outlined in Table 2004. A total of 89% of the CLP cases (n = 1402) and 79% of CPO cases (n = 631) were isolated. Any use of corticosteroids four weeks prior through 12 weeks after conception was reported by mothers of 35 (2.3%) infants with CLP (OR, 1.0; 95% CI, 0.7–1.4) and mothers of 13 (1.7%) infants with CPO (OR, 0.7; 95% CI, 0.4–1.2), and by mothers of 137 (2.4%) control infants (Table 1999). There was no association by route of administration (systemic, nasal/inhaled, topical or other use) or specific components of corticosteroids (Prednisone, Beclomethasone, Budesonide, Fluticasone, Triamcinolone). Furthermore, we did not find associations at more specific time windows of exposure (Table 2007).

By combining the earlier data with more recent data, the total cohort included mothers of 2731 infants with CLP, 1429 infants with CPO, and 10063 controls, delivered from October 1997 through December 2009. Mothers of 69 (2.6%) infants with CLP (OR 1.2 95% CI, 0.9–1.6), 19 (1.3%) infants with CPO (OR 0.6, 95% CI, 0.4–1.0) and 214 (2.1%) controls reported using any corticosteroids from 4 weeks before 12 weeks after gestation (Table 1999). We did not find an association by route of administration or component of corticosteroid in the combined data, with the possible exception of prednisone (OR, 1.9; 95% CI, 1.0–3.7) (Table 1999). Results by time window of exposure were inconsistent (Table 2007). For CLP, odds ratios ranged from 2.8 (95% CI, 1.3–5.9) for exposures only during week 1 to 4 and 5 to 8 after conception to 0.5 (95% CI, 0.1–1.6) for exposures during weeks 9 to 12.

For analyses of any corticosteroid use we adjusted for maternal race-ethnicity (Non-Hispanic white, Non-Hispanic black, Hispanic, Other, and unknown), education (<High school graduation, High school graduation, 1–3 years of college, ≥ 4years of college and unknown), intake of folic acid (any, none, and unknown), smoking (any [active], none, unknown), and study center (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, New York, North Carolina, Texas, and Utah), and we excluded nonisolated cases in the pooled data. We conducted additional analyses restricting to states that participated in the study for the whole time period (Arkansas, California, Georgia, Iowa, Massachusetts, New York, and Texas). These modifications did not appreciably change our estimates. Length of time to interview was slightly shorter for mothers reporting corticosteroid use. This was true for both cases (mean time 9.2 months for mothers reporting corticosteroid use and 10.5 for no use) and controls (mean time 8.2 months for mothers reporting corticosteroid use and 9.1 for no use).

The main results from the two time periods (1997–2002 vs. 2003–2009) are illustrated in Figure 1.

Recent data from the NBDPS provided no support for an association between maternal corticosteroid use during early pregnancy and delivering an infant with an orofacial cleft. This is in contrast to results from the first 6 years of the NBDPS, for which there was an association with CLP but not CPO (Carmichael et al., 2007). The component of corticosteroid most strongly associated with delivering an infant with CLP in the data from 1997 to 2002 was systemic prednisone; OR 2.7 (95% CI 1.1–6.7). This association was much weaker in the data from 2003 to 2009; OR 1.4 (95% CI 0.6–3.6).

When comparing corticosteroid use between the early and more recent data (1997–2002 vs. 2003–2009) there was increased use among controls (1.7–2.4%) and CPO cases (1.0–1.7%), however, there was a decrease among CLP cases (2.9 to 2.3%). This resulted in weaker associations with CLP in the more recent data. We are not aware of any significant changes in the study protocol, case ascertainment, or recruitment of cases or controls that could explain these differences. Given the small numbers, it is not possible to determine whether the frequency of reported use represents a trend or lies within the range of normal variation. We therefore suggest that our best estimates of the association of corticosteroids and orofacial clefts in the NBDPS data are those derived from the pooled data. Participation rates declined over the two time periods, from 70% to 63% overall. While this decline is substantial, participation is still in a range usually regarded as acceptable for observational studies.

The strongest association by component observed in the pooled data was for CLP and prednisone (OR, 1.9; 95% CI, 1.0–3.7). The number of control mothers reporting use of prednisone was stable at 0.3% during both time periods (1997–2002 and 2003–2009), while the proportion of case mothers (CLP) reporting prednisone went down from 0.7% to 0.4%. Given the substantial difference between the association in the earlier and later data (ORs of 2.7 vs. 1.4), and its marginal statistical significance, we recommend interpreting this result with caution. This also applies to associations between corticosteroid exposures by specific time period, because they were so variable. The strongest finding was for exposures during week 1 to 4 and 5 to 8 after conception (OR, 2.8; 95% CI, 1.3, 5.9, for the pooled data). The OR in the early data was 7.3 (95% CI, 1.8–29.4) and in the later data 1.9 (95% CI, 0.7–5.0). Although we recommend interpreting these results with caution, an association by timing of exposure cannot be dismissed completely. For comparison, a large US study reported that 0.8% of women received first trimester prescriptions for systemic corticosteroids (Andrade et al., 2004), with actual medication use probably less than 100% (Olesen et al., 2001). Furthermore, the prevalence is similar to what has been found in the National Health and Nutrition Examination Study (NHANES) during the years 1999 to 2008, where 0.5% of women aged 20 to 29 years and 0.6% of women aged 30 to 39 years reported use of oral corticosteroids (Overman et al., 2013). The NHANES data also indicate a trend toward lower prevalence of oral corticosteroid use from 1999 to 2008.

The earliest report of corticosteroids causing clefts was a study in mice (Fraser and Fainstat, 1951). Since then, studies have shown that corticosteroids are involved in cellular processes that lead to fusion of the palatal shelves, which can be disrupted by altering physiological corticosteroid levels (Pratt and Salomon, 1980; Piddington et al., 1983; Ziejewski et al., 2012). Studies have shown that teratogenicity can vary across species (Nau, 1986). Such studies have involved systemic corticosteroids, at doses that are 15 to 150 times human doses; thus, their comparability to the human condition is uncertain.

Previous epidemiological studies on corticosteroid use during early pregnancy and the risk of delivering an infant with an orofacial cleft are outlined in Table 2013. Systemic corticosteroid use in early pregnancy has been associated with delivering an infant with CLP in some previous epidemiological studies in humans (Czeizel and Rockenbauer, 1997; Rodríguez-Pinilla and Luisa Martínez-Frías, 1998; Carmichael and Shaw, 1999; Pradat et al., 2003; Carmichael et al., 2007), one of which also reported an association with CPO (Carmichael and Shaw, 1999). Studies from Denmark, Norway, Sweden have found no association with systemic use in early pregnancy and orofacial clefts in the offspring, and a weak association with dermatological corticosteroids (Källén, 2003; Hviid and Mølgaard-Nielsen, 2011; Skuladottir et al., 2013). However, a recent population-based cohort study from the UK did not find an association between dermatological corticosteroids and clefts (Chi et al., 2013). In sum, the current literature is inconsistent regarding the association of first-trimester corticosteroid use and orofacial clefts in humans. The previous studies are limited by sample size, with number of cases (CLP and CPO combined) ranging from 8 to 1232.

The NBDPS included 4072 pregnancies resulting in either CLP or CPO, with 23 (0.6%) mothers reporting systemic corticosteroid use, making it the largest study exploring this potential association to date. Other strengths include the population-based design and the detailed assessment on corticosteroid mode, specific component used and the detailed time windows of exposure. We lacked information on dose and indication, which were limitations. Other potential limitations include recall bias (mean time to interview was slightly shorter for the mothers who reported corticosteroid use than the mothers who did not report use) and selection bias (participation was 72% for case mothers and 65% for control mothers). In the NBDPS questionnaire, there is no specific question for dermatological disease or treatment, and dermatological medication is consequently underreported and estimates are therefore inaccurate. Under-reported use of other types of corticosteroids is also possible but difficult to determine.

Maternal use of corticosteroids is not associated with delivering an infant with an orofacial cleft in the NBDPS. This analysis is consistent with recent results from large population-based studies (Källén, 2003; Hviid and Mølgaard-Nielsen, 2011; Skuladottir et al., 2013). These data help to inform the clinical risk-benefit decision for use of corticosteroids during the first trimester of pregnancy.