Objective
Prior studies have reported increased risks of congenital heart defects (CHD) and pyloric stenosis (PS) after prenatal exposure to macrolide antibiotics. We sought to assess the association between maternal use of erythromycin and nonerythromycin macrolides and the risks of CHD and PS.
Study Design
Among participants in the Slone Epidemiology Center Birth Defects Study from 1994 through 2008, we identified 4132 infants with CHD and 735 with PS as cases, and 6952 infants without any malformation as controls. We estimated odds ratios (ORs) and 95% confidence intervals (CIs) associated with use of erythromycin or nonerythromycin macrolides in each trimester using conditional logistic regression and adjusting for risk factors for CHD and PS, fever, specific types of infections, and their associated treatments.
Results
During the first trimester, 0.4% and 0.7% of control women had used erythromycin and nonerythromycin macrolides, respectively. Compared to non-use during pregnancy, first-trimester exposure to erythromycin was not associated with an increased risk of CHD (OR, 1.3; 95% CI, 0.6−2.6) or PS (OR, 0.9; 95% CI, 0.3−3.0). The corresponding ORs for nonerythromycin macrolides were 0.7 (95% CI, 0.4−1.3) for CHD and 1.7 (95% CI, 0.6−4.6) for PS. We found no association between third-trimester exposure to erythromycin or nonerythromycin macrolides and the risk of PS. Hypothesis generation analyses did not identify appreciable associations between maternal use of macrolides and other common specific birth defects.
Conclusion
We found no meaningful associations between the risks of CHD, PS, and other common malformations in relation to use of macrolides in pregnancy.
Macrolide antibiotics are frequently used for presumed or documented gram-positive lower and upper respiratory infections, soft tissue infections, and Helicobacter pylori −related peptic ulcer. Moreover, for the treatment of Chlamydia and other selected infections in pregnancy, erythromycin has been specifically recommended because other effective antibiotics for such infections are contraindicated for pregnant women. Because these indications are common and azithromycin and erythromycin are classified as Food and Drug Administration category B, macrolide antibiotics are the second most frequently used antibacterial class during pregnancy in the United States. Macrolide antibiotics are often subdivided into erythromycin, the first-introduced macrolide, and nonerythromycin drugs, including clarithromycin and azithromycin, which have fewer effects on gastrointestinal motility than erythromycin.
Exposure to erythromycin in early pregnancy has been associated with an increased risk of congenital heart defects (CHD) (odds ratio [OR], 1.84; 95% confidence interval [CI], 1.29−2.62); the association was largely attributed to unspecified CHD (OR, 3.57; 95% CI, 1.70−6.12). However, Cooper et al did not find an increase in the risk of CHD after exposure to either erythromycin or nonerythromycin macrolides. The National Birth Defects Prevention Study also published null findings for macrolides overall in relation to CHD, but they did not differentiate erythromycin from nonerythromycin macrolides.
In 1999, infantile hypertrophic pyloric stenosis (PS) was linked to exposure to macrolide antibiotics in postnatal days 2-17, a finding confirmed in 2001. Kallen et al reported an elevated risk of PS among the offspring of women who took erythromycin in early pregnancy (risk ratio, 3.03; 95% CI, 1.08−8.50). Cooper et al did not replicate this association, but found an elevated risk of PS associated with exposure to nonerythromycin macrolides prescribed any time during pregnancy (OR, 2.77; 95% CI, 1.22−6.30). Using data from the Slone Birth Defects Study (BDS) (1976 through 1998), Louik et al found no association between PS and exposure to either type of macrolide antibiotics >32nd gestational week (OR, 0.7; 95% CI, 0.3−1.8 for erythromycin and OR, 1.1; 95% CI, 0.3−3.6 for nonerythromycin macrolides), but there were very few subjects exposed to nonerythromycin macrolides in this study. These findings have raised concerns regarding maternal use of macrolide antibiotics in either early or late pregnancy.
Despite being commonly prescribed during pregnancy, the safety profile of macrolide antibiotics is yet to be determined, not only with regard to the hypothesized risks of CHD and PS, but also with regard to the range of other specific major birth defects. We therefore sought to test the hypotheses that the risks of CHD and PS are elevated among infants or fetuses exposed to erythromycin and/or nonerythromycin macrolides during pregnancy and, in exploratory analyses, to identify possible associations with other specific defects. The analyses utilized data from the BDS, an ongoing program of case-control surveillance of medications in relation to birth defects.
Materials and Methods
Study population
The BDS was established in 1976, and since that time has interviewed mothers of malformed infants ascertained through review of admissions and discharges at major referral hospitals and clinics in the greater metropolitan areas of Boston, Philadelphia, Toronto, and San Diego, and through statewide birth defects registries in New York State (since 2004) and Massachusetts (since 1998). For hospital-based surveillance, the subjects’ physicians are asked to confirm the diagnosis and mothers are asked to provide medical record releases to permit confirmation of the infant’s condition. Infants with isolated minor defects are excluded. Beginning in 1992, the BDS also enrolled a sample of mothers of nonmalformed infants as controls: initially these infants were identified exclusively at study hospitals but, since 1998, the BDS also includes a population-based random sample of newborns in Massachusetts. The study has been approved by the Boston University Institutional Review Board and the institutional review boards of all relevant participating institutions. It is fully compliant with the requirements of the Health Insurance Portability and Accountability Act. The current analysis was restricted to women interviewed from 1994 through 2008 because full ascertainment of our control group, nonmalformed infants, was not underway until 1994. Among eligible subjects in the last decade, the mothers of 73% of malformed infants and 68% of nonmalformed controls contacted agreed to an interview and provided informed consent.
Cases
Hypothesis testing
Cases consisted of 4132 infants and fetuses with a diagnosis of CHD and 735 infants with PS. We excluded from analysis infants with chromosomal defects, known mendelian inherited disorders, syndromes, DiGeorge sequence (associated with 22q deletion), and metabolic and functional disorders. CHD or PS complicated with other defects (but not as part of an identified chromosomal or mendelian inherited syndrome) were included in the general analysis and studied separately in a secondary analysis.
Exploratory analyses
Other major defects examined included the following categories: 1348 oral clefts, 1138 central nervous system defects, 308 respiratory system defects, 1825 gastrointestinal system defects, 1099 genital system defects, 1511 urinary system defects, 1948 musculoskeletal system defects, and 385 others. Where there were sufficient numbers of subjects with specific defects, those were considered as well. The same exclusion criteria described above also applied to these case groups.
Controls
Our control group consisted of 6952 infants without any malformation.
Interviews
Within 6 months of the subject’s delivery, trained study nurses unaware of study hypotheses interview mothers of study subjects. The 45- to 60-minute interview is detailed and structured and includes questions on maternal demographic characteristics, mothers’ medical histories, obstetric histories, maternal health behaviors and occupation, and a detailed history of the use of medication (including prescription, over-the-counter, and vitamin and herbal products) from 2 months before the date of the last menstrual period (LMP) through the entire pregnancy. Recall of medication exposures is enhanced by questions regarding indications for use (eg, infections), and a list of specifically named medications, which includes, among other antibiotics, erythromycin and nonerythromycin macrolides.
Mothers who report taking a particular medication are further asked to identify, as accurately as possible, the dates when use began and ended. Recall of the timing of medication use is enhanced by the use of a calendar that highlights the mother’s reported LMP date and her delivery date. Further, subjects are asked how certain they are about each of these dates. Interviewers record the certainty of each reported date as follows: (1) exact, if the exact date is reported, (2) estimated, if a date is stated as an estimate, or (3) sometime in a given month, if the day within a month is unknown. Mothers who cannot recall the month are considered to have unknown dates of exposure. Mothers are also asked details about their pattern of use of the particular medication, including duration (days of treatment), frequency of use (eg, days per week or month), and specific doses. We defined exposure as systemic use of erythromycin or nonerythromycin macrolides.
Algorithm to classify timing of exposure
We developed an exposure classification algorithm taking into account recall uncertainty in reported timing of medication exposure. For uncertain start/stop dates reported as being sometime in a month, we considered the possible exposure period to be the widest interval consistent with that report (eg, if a mother reported medication use sometime in May, we assigned May 1 as her start date and May 31 as her stop date). When a mother is not certain about the exact exposure date, assuming different start dates consistent with the reported range, the exposure period may overlap with the window of interest or not. If the exposure period includes window of interest with any possible start dates of the exposure, we classified the mother as “likely exposed.” Details about all possible scenarios were previously published. To minimize misclassification of exposure, only women classified as “likely exposed” to each medication comprised our exposure group of primary interest.
Data analysis
For both hypothesis testing and the exploratory analyses, we used conditional logistic regression to estimate ORs and CIs for exposure to macrolide antibiotics in each trimester and the risk of specific birth defects. For the timing of pregnancy periods, we defined the LMP date based on early ultrasound examination; in the absence of that information, we relied on maternal report. We defined the estimated date of conception (EDC) as 14 days after the LMP date and the first trimester of pregnancy as 90 days following the EDC (encompassing the etiologically important period of vulnerability to the development of most structural birth defects). Since that period is not established for PS, we considered exposure in each trimester for that defect. The second and third trimesters were defined as 91st-180th day and 181st day to the end of the pregnancy, respectively, following the EDC. It was postulated that erythromycin exposure might cause PS by interacting with motilin receptors, which are present in the fetus beginning at 32 weeks’ gestation. Therefore, in a sensitivity analysis for PS, we also examined exposure to erythromycin >32nd gestational week. We used multivariate conditional logistic regression to adjust for the following factors: (1) geographic area of participants’ residences and calendar year when they were ascertained (to account for secular trends and regional variations in use of antimicrobials as well as recruitment of cases with specific malformations); (2) risk factors for congenital malformations, including maternal age, race, education level, prepregnancy body mass index, family history of CHD or PS, diabetes mellitus, first-trimester cigarette smoking, periconceptional folic acid supplement, and multiple pregnancy; (3) infection-related factors, including urinary tract, respiratory, or vaginal/yeast infection, sexually transmitted disease, other kinds of infection, and/or febrile events with/without treatment that occurred in the first trimester. Further adjustment for use of anticonvulsants, alcohol consumption in the first trimester, and history of infertility with/without treatment did not change our results appreciably. The reference category was no exposure to the medication of interest at any time and in any form (systemic or topical use) from 56 days before the LMP date through the end of pregnancy. To avoid unstable estimates, ORs with <5 exposed cases or 100 total cases were not calculated. Statistical analyses utilized SAS 9.2 (SAS Institute Inc, Cary, NC).
Sensitivity analyses
Because preterm deliveries, stillbirths, and terminations may occur differentially among cases vs nonmalformed controls and because they would have a shorter opportunity for exposure late in pregnancy, we conducted a sensitivity analysis that excluded these babies in the analysis for late exposure to macrolides (in the third trimester or >32nd gestational week).
If a mother reported having used an antibiotic but the type was not otherwise specified (NOS), she would be considered a user of an antibacterial but not a user of a specific antibiotic. Therefore, the reference group for macrolide antibiotics use could include a few users of other antibiotics, including antibiotic NOS, some of whom might have actually used a macrolide antibiotic. In a sensitivity analysis, we excluded users of antibiotic NOS from the reference group.
Results
Among the control group, 9.5% of women reported having used at least 1 antibacterial in the first trimester, the most common type being penicillins (3.4%), followed by macrolide antibiotics (1.1% overall; 0.4% for erythromycin and 0.7% for nonerythromycin macrolides); 4.1% reported having taken an antibiotic NOS. From 1994 through 2008, there was a decreasing trend in erythromycin use and an increasing trend in use of nonerythromycin macrolides during pregnancy ( Figure ).
Hypothesis testing: CHD and PS
The distributions of selected characteristics are shown for cases and controls in Table 1 . Geographic area of participants’ residences and calendar time were highly associated with the relative frequency of CHD and PS cases because of secular trends and regional variations in recruitment of study subjects with different malformations (data not shown). Therefore all the estimates presented were adjusted for both factors.
| Variables | Levels | CHD cases (N = 4132) | PS cases (N = 735) | Controls (N = 6952), n (%) | ||
|---|---|---|---|---|---|---|
| n (%) | OR (95% CI) a | n (%) | OR (95% CI) a | |||
| Mother’s age, y | <20 | 238 (5.8) | 1.2 (0.9–1.4) | 37 (5.0) | 1.2 (0.8–1.7) | 394 (5.7) |
| 20-24 | 627 (15.2) | 1.2 (1.1–1.4) | 122 (16.6) | 1.5 (1.2–1.9) | 888 (12.8) | |
| 25-29 | 1098 (26.6) | 1.0 (0.9–1.1) | 194 (26.4) | 1.2 (0.9–1.4) | 1703 (24.5) | |
| 30-34 | 1341 (32.5) | 1.0 (Ref) | 241 (32.8) | 1.0 (Ref) | 2460 (35.4) | |
| 35-39 | 679 (16.4) | 1.1 (0.9–1.2) | 121 (16.5) | 1.1 (0.9–1.4) | 1254 (18.0) | |
| ≥40 | 149 (3.6) | 1.2 (0.9–1.5) | 20 (2.7) | 1.1 (0.7–1.8) | 253 (3.6) | |
| Mother’s race/ethnicity | White | 2890 (69.9) | 1.0 (Ref) | 618 (84.1) | 1.0 (Ref) | 5144 (74.0) |
| Black | 299 (7.2) | 0.9 (0.8–1.1) | 32 (4.4) | 0.5 (0.3–0.7) | 489 (7.0) | |
| Hispanic | 594 (14.4) | 1.0 (0.8–1.2) | 50 (6.8) | 0.9 (0.7–1.3) | 844 (12.1) | |
| Asian, Pacific Islander | 214 (5.2) | 1.4 (1.2–1.6) | 23 (3.1) | 0.5 (0.3–0.7) | 336 (4.8) | |
| Others | 135 (3.3) | 1.8 (1.4–2.5) | 12 (1.6) | 0.6 (0.3–1.1) | 139 (2.0) | |
| Mother’s education | <13 | 1395 (33.8) | 1.0 (Ref) | 267 (36.3) | 1.0 (Ref) | 1887 (27.1) |
| 13-15 | 1034 (25.0) | 0.8 (0.7–0.9) | 194 (26.4) | 0.7 (0.6–0.9) | 1699 (24.4) | |
| ≥16 | 1699 (41.1) | 0.7 (0.6–0.8) | 274 (37.3) | 0.6 (0.5–0.7) | 3363 (48.4) | |
| Mother’s prepregnancy BMI | <18.5 | 232 (5.6) | 0.9 (0.8–1.1) | 45 (6.1) | 0.9 (0.7–1.3) | 423 (6.1) |
| 18.5-24.9 | 2334 (56.5) | 1.0 (Ref) | 432 (58.8) | 1.0 (Ref) | 4340 (62.4) | |
| 25-29.9 | 865 (20.9) | 1.1 (1.0–1.3) | 145 (19.7) | 1.1 (0.9–1.4) | 1334 (19.2) | |
| 30-34.9 | 369 (8.9) | 1.5 (1.2–1.7) | 67 (9.1) | 1.6 (1.2–2.1) | 475 (6.8) | |
| 35-39.9 | 143 (3.5) | 1.5 (1.2–2.0) | 30 (4.1) | 2.2 (1.4–3.4) | 167 (2.4) | |
| ≥40 | 79 (1.9) | 1.9 (1.4–2.6) | 6 (0.8) | 0.9 (0.4–2.3) | 91 (1.3) | |
| Diabetes mellitus | None | 3747 (90.7) | 1.0 (Ref) | 698 (95.0) | 1.0 (Ref) | 6618 (95.2) |
| Diagnosed before pregnancy | 101 (2.4) | 5.6 (3.6–8.7) | 4 (0.5) | 1.4 (0.5–4.1) | 37 (0.5) | |
| Diagnosed during pregnancy | 280 (6.8) | 1.8 (1.5–2.2) | 33 (4.5) | 1.4 (0.9–2.0) | 290 (4.2) | |
| Folic acid supplementation | None | 166 (4.0) | 1.0 (Ref) | 42 (5.7) | 1.0 (Ref) | 248 (3.6) |
| Any periconceptional use b | 2434 (58.9) | 1.0 (0.8–1.2) | 434 (59.0) | 0.8 (0.5–1.1) | 4431 (63.7) | |
| Family history of CHD c | None | 3469 (84.0) | 1.0 (Ref) | — | 6318 (90.9) | |
| First degree | 308 (7.5) | 2.0 (1.7–2.5) | 267 (3.8) | |||
| Second degree | 259 (6.3) | 1.9 (1.6–2.4) | 264 (3.8) | |||
| Third degree | 92 (2.2) | 1.6 (1.1–2.2) | 103 (1.5) | |||
| Family history of PS c | None | — | 644 (87.6) | 1.0 (Ref) | 6909 (99.4) | |
| First degree | 58 (7.9) | 24.2 (13.6–42.8) | 24 (0.3) | |||
| Second degree | 23 (3.1) | 14.4 (6.7–30.8) | 15 (0.2) | |||
| Third degree | 10 (1.4) | 12.2 (3.6–41.5) | 4 (0.1) | |||
| Primiparous | No | 2860 (69.2) | 1.0 (Ref) | 506 (68.8) | 1.0 (Ref) | 4833 (69.5) |
| Yes | 1272 (30.8) | 1.0 (0.9–1.1) | 229 (31.2) | 1.0 (0.8–1.2) | 2119 (30.5) | |
| Multiple pregnancy | Single | 3908 (94.6) | 1.0 (Ref) | 698 (95.0) | 1.0 (Ref) | 6749 (97.1) |
| Multiple | 224 (5.4) | 2.5 (2.0–3.2) | 37 (5.0) | 2.1 (1.4–3.2) | 203 (2.9) | |
| Smoking in first trimester d | Nonsmoker | 2376 (57.5) | 1.0 (Ref) | 351 (47.8) | 1.0 (Ref) | 4034 (58.0) |
| Smoked during first trimester | 830 (20.1) | 1.2 (1.0–1.3) | 217 (29.5) | 1.7 (1.4–2.1) | 1207 (17.4) | |
| Ex-smoker/smoke outside first trimester | 926 (22.4) | 1.0 (0.9–1.1) | 167 (22.7) | 1.1 (0.9–1.4) | 1711 (24.6) | |
| Infection in first trimester d , e | UTI | 199 (4.8) | 1.2 (1.0–1.5) | 33 (4.5) | 1.1 (0.8–1.7) | 272 (3.9) |
| Respiratory infection | 908 (22.0) | 1.2 (1.0–1.3) | 158 (21.5) | 1.1 (0.9–1.3) | 1403 (20.2) | |
| Yeast/vaginal | 255 (6.2) | 1.0 (0.8–1.2) | 57 (7.8) | 1.1 (0.8–1.5) | 421 (6.1) | |
| STD | 77 (1.9) | 1.0 (0.7–1.4) | 11 (1.5) | 1.0 (0.5–1.9) | 130 (1.9) | |
| Others | 313 (7.6) | 1.0 (0.9–1.2) | 74 (10.1) | 1.1 (0.9–1.5) | 510 (7.3) | |
| Febrile event in first trimester d | None | 3313 (80.2) | 1.0 (Ref) | 574 (78.1) | 1.0 (Ref) | 5601 (80.6) |
| Fever ≥101°F in first trimester | 97 (2.3) | 1.3 (1.0–1.8) | 19 (2.6) | 1.7 (1.0–2.8) | 130 (1.9) | |
| Fever <101°F in first trimester | 121 (2.9) | 1.1 (0.8–1.4) | 23 (3.1) | 1.1 (0.7–1.7) | 201 (2.9) | |
a Based on conditional logistic regression, adjusted for region of participants’ residences and calendar year of ascertainment;
b Any folic acid supplementation during −1∼+1 month around last menstrual period;
c Family history of CHD when infants with CHD served as cases; and family history of PS when infants with PS served as cases;
d First trimester of pregnancy was defined as 90 d after estimated date of conception (14 d after last menstrual period);
e Ref for each type of infection was no infection under comparison in first trimester.
As shown in Table 2 , we found no association between first-trimester maternal use of macrolides as a class and the risk of CHD (OR, 0.9; 95% CI, 0.6−1.3) or PS (OR, 1.3; 95% CI, 0.6−2.8). For erythromycin, the risk of CHD was 1.3 (95% CI, 0.6−2.6) and for PS it was 0.9 (95% CI, 0.3−3.0). The corresponding ORs for nonerythromycin macrolides were 0.7 (95% CI, 0.4−1.3) for CHD and 1.7 (95% CI, 0.6−4.6) for PS. We also found no meaningful association between the risk of PS and exposure to macrolides overall or to erythromycin or nonerythromycin macrolides in the second and third trimesters ( Table 2 ). Restriction to data after 1998 (excluding the data overlapping with our prior study of PS ) did not materially affect our results (data not shown). Examining specific cardiac defects within the overall CHD category, we also observed no meaningful associations. For example, the OR of septal defects was 1.3 (95% CI, 0.6−3.0) for erythromycin and 0.8 (95% CI, 0.4−1.4) for nonerythromycin macrolides ( Table 3 ). In the sensitivity analysis for etiologically relevant timing of exposure for PS, no associations were found with either erythromycin or nonerythromycin macrolides >32nd gestational week. Excluding preterm deliveries, stillbirths, and therapeutic abortions did not appreciably affect our risk estimates for late-pregnancy exposure to macrolide antibiotics or to the macrolide subgroups (data not shown). Exclusion of users of antibiotic NOS from the analysis or restriction to isolated CHD and PS (without other major malformations) also did not materially affect the results (data not shown).
| Variable | Nonmalformed controls (N = 6952), n (%) | Cases of CHD (N = 4132) | Cases of PS (N = 735) | ||
|---|---|---|---|---|---|
| n (%) | OR (95% CI) a | n (%) | OR (95% CI) a | ||
| Any macrolides | |||||
| No exposure during pregnancy | 6655 (95.7) | 3948 (95.5) | 1.0 (Ref) | 693 (94.3) | 1.0 (Ref) |
| Exposure in first trimester b | 70 (1.0) | 46 (1.1) | 0.9 (0.6–1.3) | 12 (1.6) | 1.3 (0.6–2.8) |
| Exposure in second trimester b | 74 (1.1) | 47 (1.1) | 1.1 (0.7–1.7) | 7 (1.0) | 1.3 (0.5–3.0) |
| Exposure in third trimester b , c | 71 (1.0) | 47 (1.1) | 1.0 (0.6–1.6) | 11 (1.5) | 1.3 (0.6–2.9) |
| Erythromycin | |||||
| No exposure during pregnancy | 6828 (98.2) | 4064 (98.4) | 1.0 (Ref) | 717 (97.6) | 1.0 (Ref) |
| Exposure in first trimester d | 28 (0.4) | 18 (0.4) | 1.3 (0.6–2.6) | 4 (0.5) | 0.9 (0.3–3.0) |
| Exposure in second trimester d | 27 (0.4) | 15 (0.4) | 0.9 (0.4–2.0) | 4 (0.5) | 1.5 (0.4–4.8) |
| Exposure in third trimester c , d | 20 (0.3) | 15 (0.4) | 1.1 (0.5–2.6) | 5 (0.7) | 1.5 (0.5–5.1) |
| Nonerythromycin macrolides | |||||
| No exposure during pregnancy | 6773 (97.4) | 4013 (97.1) | 1.0 (Ref) | 711 (96.7) | 1.0 (Ref) |
| Exposure in first trimester e | 43 (0.6) | 29 (0.7) | 0.7 (0.4–1.3) | 8 (1.1) | 1.7 (0.6–4.6) |
| Exposure in second trimester e | 48 (0.7) | 32 (0.8) | 1.2 (0.7–2.0) | 3 (0.4) | — |
| Exposure in third trimester c , e | 51 (0.7) | 32 (0.8) | 1.0 (0.6–1.7) | 7 (1.0) | 1.5 (0.6–3.8) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
