OBJECTIVE: To assess the association between prenatal antibiotics, including erythromycin, and infantile hypertrophic pyloric stenosis in a large cohort of infants.
METHODS: This was a retrospective cohort study of births to women enrolled in Tennessee Medicaid/TennCare, 1985–1997. Prescriptions for erythromycin, nonerythromycin macrolides, and other antibiotics were identified from pharmacy files linked with birth certificate files. The primary study outcome was development of pyloric stenosis in the infant, identified from linked hospital discharge diagnosis and surgical procedure codes.
RESULTS: The cohort included 260,799 mother/infant pairs. Among these women, 13,146 filled prescriptions for erythromycin (50.4 per 1000), and 621 filled prescriptions for nonerythromycin macrolides (2.4 per 1000). There was no association with prenatal erythromycin prescription and infantile hypertrophic pyloric stenosis either after 32 weeks' gestation (adjusted odds ratio 1.17, 95% confidence interval, 0.84, 1.64, P = .33) or at any time during pregnancy (adjusted odds ratio 1.15, 95% confidence interval 0.84, 1.56, P = .36). There was an association between maternal prescriptions for nonerythromycin macrolides and infantile hypertrophic pyloric stenosis (adjusted odds ratio 2.77, 95% confidence interval 1.22, 6.30, P = .01).
CONCLUSION: The hypothesized association between erythromycin and infantile pyloric stenosis was not seen. Causal inference from the association between prenatal nonerythromycin macrolides and infantile hypertrophic pyloric stenosis is limited by the small number of affected children and the evidence of other differences between users of nonerythromycin macrolides and controls.
Prenatal prescriptions for erythromycin were not associated with pyloric stenosis; an unanticipated association was found between prenatal nonerythromycin macrolides and pyloric stenosis.
From the Division of General Pediatrics, Department of Pediatrics, and Division of Pharmacoepidemiology, Department of Preventive Medicine, Vanderbilt University, Nashville, Tennessee.
Address reprint requests to: William O. Cooper, MD, MPH, Vanderbilt University School of Medicine, Division of General Pediatrics, Suite 5028 MCE, Nashville, TN 37232–8555; E‐mail: firstname.lastname@example.org.
Dr. Cooper received support from the Generalist Physician Faculty Scholars Program of the Robert Wood Johnson Foundation (#03816). Drs. Ray and Griffin received support from the Centers for Education and Research in Therapeutics (CERT) program of the Agency for Healthcare Research and Quality (#1U18HS10384–01).
Received October 25, 2001. Received in revised form January 15, 2002. Accepted February 14, 2002.
A series of reports1–4 and a recent population‐based study5 have suggested an association between early exposure to erythromycin and the development of infantile hypertrophic pyloric stenosis in young infants. Infants developing pyloric stenosis after erythromycin exposure have earlier onset of symptoms than typical cases of pyloric stenosis,1–6 suggesting possible differences in pathophysiology for erythromycin‐related pyloric stenosis.
One theory of erythromycin exposure causing pyloric stenosis postulates that erythromycin interacts with motilin receptors, inducing strong gastric and pyloric bulb contractions and resulting in pylorus hypertrophy.7,8 Motilin receptors are present in the fetus beginning at 32 weeks' gestation.7 Thus, it is plausible that maternal use of erythromycin after 32 weeks' gestation could cause hypertrophy of the fetal pylorus and subsequent postnatal infantile pyloric stenosis. Other antibiotics resulting in increased gastric motility could plausibly be related to pyloric stenosis as well. There are no other clearly understood etiologies for pyloric stenosis.
The current study was designed to estimate the association between prenatal prescriptions for erythromycin and other antibiotics and infantile hypertrophic pyloric stenosis (IHPS) in a large cohort of infants.
MATERIALS AND METHODS
Mother/infant pairs were included in this retrospective cohort study if the birth occurred in Tennessee between 1985 and 1997 and information in the Tennessee birth certificate files was complete. To allow for identification of prenatal antibiotic use and postnatal development of pyloric stenosis, both mother and infant were required to be enrolled in Medicaid (1985–1993) or TennCare (Tennessee's managed care program for Medicaid enrollees and uninsured individuals, 1994–1997) at the time of delivery. Because the Tennessee Medicaid program had already expanded coverage to pregnant women before the 1994 initiation of TennCare, there was no difference in the proportion of Tennessee births covered by Medicaid/TennCare before and after 1994.9 Only full‐term infants were included in the study because pyloric stenosis is rare in preterm children and the proposed mechanism of erythromycin‐induced pyloric stenosis involves motilin receptors not present until late in pregnancy. Children who filled erythromycin prescriptions in the first 14 days of life were excluded because of the previously described relationship between early infant exposure to erythromycin and pyloric stenosis.1–5 Because drugs received in the hospital are not included in the databases used for this study, children with prolonged neonatal intensive care unit stays were excluded. Children with prolonged neonatal intensive care unit stays were defined as children who were discharged from a birth hospital after 14 days of life. Mother/infant pairs where the infant died in the first 90 days of life were also excluded. To ensure complete ascertainment of records, mothers were required to be continuously enrolled during the last trimester of pregnancy, and infants were required to be continuously enrolled from birth through 90 days of life (86–90% of infants).10
Maternal antibiotic prescriptions during pregnancy were identified by linking birth certificate information with Medicaid/TennCare pharmacy files. Prescriptions were identified from the date of the last menstrual period recorded in the birth certificate file through the date of delivery. Antibiotics included erythromycin, nonerythromycin macrolide antibiotics (lincomycin, clindamycin, clarithromycin, azithromycin, and dirithromycin), and other nonmacrolide antibiotics (cephalosporins, fluoroquinilones, penicillins, sulfa drugs, tetracyclines, and chloramphenicol). Nonerythromycin macrolide antibiotics were considered separately from erythromycin because they do not interact with motilin receptors as strongly as erythromycin.11 Prescriptions were grouped by the week of gestation, with particular emphasis on prescriptions from 32 weeks' gestation to delivery. Categories allowed for multiple antibiotic prescriptions for a pregnancy. If a mother filled a prescription for the same antibiotic twice during the same pregnancy, the latest prescription was included because it was postulated that late prescriptions would most plausibly result in increased risk. If a mother filled prescriptions for both erythromycin and a nonerythromycin macrolide, priority was given to the erythromycin because the primary study hypothesis involved exposure to erythromycin.
Medicaid/TennCare encounter files were searched to identify infants having pyloric stenosis hospitalizations between 3 and 90 days of life. Restricting pyloric stenosis hospitalizations to those occurring from 3 days of life on allowed for varying hospital lengths of stay after delivery during the study period, to give all infants in the base population equal opportunity to experience the study outcome. Initial screening included the specific International Classification of Diseases, 9th Revision, Clinical Modification12 (ICD‐9‐CM) code for pyloric stenosis (750.5) and ICD‐9‐CM codes for other diagnoses, which could be coded for pyloric stenosis (acquired pyloric stenosis [ICD‐9‐CM code 537.0] and pylorospasm [ICD‐9‐CM code 537.81]). Procedure codes from physician and hospital claims were searched to identify codes for pyloromyotomy, the definitive surgical procedure for pyloric stenosis (Current Procedural Terminology13 code 43520 or ICD‐9‐CM procedure code 43.3). Cases were defined as infants with a discharge diagnosis code for pyloric stenosis along with a procedure code for pyloromyotomy.
Strata were defined by study covariates, and the proportion of mother/infant pairs where the infant was defined as having pyloric stenosis was compared across strata. Multivariable logistic regression (SAS 8.1, SAS Institute Inc., Cary, NC) was used to control for study covariates including maternal age, maternal education, maternal residence in rural county, use of other antibiotics, infant sex, infant race, birth order, and year of birth. Maternal sociodemographic variables were included because of their relationship to antibiotic use. Infant sex, race, and birth order were included because of their relationship to the development of pyloric stenosis. Models were constructed for erythromycin use from 32 weeks' gestation to delivery and for macrolide use from 32 weeks' gestation to delivery. These models only included mothers with continuous enrollment during the third trimester of pregnancy. Models for antibiotic prescriptions at any time during pregnancy only included mothers with continuous enrollment throughout pregnancy. Prescriptions throughout pregnancy were assessed because of previous reports describing possible relationships between exposure to drugs early in pregnancy and pyloric stenosis.4,14,15
The study was reviewed and approved by the Vanderbilt University Institutional Review Board, the State of Tennessee Institutional Review Board, the Tennessee Department of Health, and the TennCare Bureau.
There were a total of 933,239 births to Tennessee residents between 1985 and 1997. Of these infants, 932,817 had complete birth certificate information (99.9%), and 400,015 (42.9% of mother/infant pairs with complete information) were enrolled in Medicaid or TennCare within 2 weeks of life. Of infants enrolled in Medicaid or TennCare, 52,437 (13.2%) had gaps in enrollment after birth, 7008 (1.8%) remained in the hospital past 14 days of life, and 1494 (0.4%) died, leaving a total of 339,076 mother/infant pairs. Of these mother/infant pairs, 303,195 (89.4%) mothers were continuously enrolled throughout the third trimester of pregnancy. Restricting the cohort further to full‐term infants (86.2%) and infants with no prescriptions for erythromycin or nonerythromycin macrolides (99.2%) after birth left a cohort of 260,799 mother/infant pairs.
Between 32 weeks' gestation and delivery, 13,146 mothers (50.4 per 1000) filled prescriptions for erythromycin (Table 1). Compared with the 247,032 mothers who did not fill prescriptions for erythromycin or macrolides, a higher proportion of mothers filling prescriptions for erythromycin were less than age 18 years, had less than 12 years of education, were black, resided in urban counties, and filled prescriptions for other antibiotics (Table 1). A smaller number of mothers (n = 621, 2.4 per 1000) filled prescriptions for nonerythromycin macrolides after 32 weeks' gestation. Compared with mothers who did not fill prescriptions for erythromycin or macrolides, a lower proportion of mothers filling prescriptions for nonerythromycin macrolides had less than 12 years of education; a higher proportion were black, resided in urban counties, and filled prescriptions for other antibiotics (Table 1).
Among mother/infant pairs meeting enrollment requirements, 679 (2.6 per 1000) infants developed pyloric stenosis. Infants with pyloric stenosis were admitted at a mean age of 39.3 days, 81.3% were male, 86.5% were white, 44.5% resided in a rural county, and 50.2% were first‐born children.
Among the 13,146 mothers with erythromycin prescriptions after 32 weeks' gestation, there were 38 cases of pyloric stenosis (2.9 per 1000) (Table 2); three cases occurred among infants born to the 621 mothers who filled prescriptions for nonerythromycin macrolides after 32 weeks' gestation (4.8 per 1000); 638 cases occurred among infants born to the 247,032 mothers who did not fill either type of prescription from 32 weeks' gestation through delivery (2.6 per 1000). In multivariable comparisons, the difference in the rates of pyloric stenosis for exposed and nonexposed infants was not statistically significant.
For the subset of mothers with continuous enrollment throughout pregnancy (n = 119,073), erythromycin use at any time during pregnancy was not associated with pyloric stenosis (Table 3). For other macrolide use at any time during pregnancy, there was an association of pyloric stenosis with maternal macrolides prescriptions, as estimated by an odds ratio of 2.77 (95% confidence interval 1.22, 6.30, P = .01). There was no difference in the proportion of younger infants in the prescription groups.
With the unexpected finding of a higher rate of pyloric stenosis in infants born to mothers filling nonerythromycin macrolide prescriptions than other groups, additional analyses were conducted. Diagnoses associated with prescriptions for nonerythromycin antibiotics and other antibiotics in study mothers were compared. There were 78% of mothers who had a health care encounter claim on the same date the prescription was filled. A higher proportion of mothers filling prescriptions for nonerythromycin macrolides had encounters with diagnostic codes indicating treatment for sexually transmitted diseases than mothers filling prescriptions for erythromycin or other antibiotics (18% versus 10%, χ2 = 124.1, P < .001).
To determine if a sexually transmitted disease or some associated factor might be responsible for the observed association between nonerythromycin macrolides and pyloric stenosis, we estimated the rate of pyloric stenosis among infants of mothers exposed to doxycyline, an antibiotic used for similar indications as the nonerythromycin macrolides. Among the 1635 pregnancies associated with doxycyline use, there were four children with pyloric stenosis (2.45 per 1000), a prevalence similar to controls (χ2 = 0.11, P = .74).
In this study of 260,799 mother/infant pairs, erythromycin prescriptions during pregnancy were not associated with an infant developing pyloric stenosis. There was, however, an association of pyloric stenosis with maternal filling of prescriptions for nonerythromycin macrolides at any time during pregnancy.
Factors that are known to increase the risk of pyloric stenosis include being first born, white, male, and having a greater birth weight.16 In addition, the seasonality of pyloric stenosis has suggested an infectious etiology.17 A series of studies examined prenatal exposure to the antinausea medication doxylamine succinate‐pyridoxine hydrochloride (Bendectin) and subsequent development of pyloric stenosis. These studies produced conflicting results.14–16 There was not strong evidence for a causal association; however, the drug was voluntarily removed from the market in 1983, partly because of the excessive costs of defending lawsuits related to Bendectin.
Infants with very early exposure to erythromycin appear to have an increased risk for pyloric stenosis and to develop pyloric stenosis at a younger age than typical cases.1,4 In addition, the prokinetic effects of erythromycin resulting in its clinical use to improve gastric emptying make it plausible that use of the drug could result in pylorus hypertrophy.18 Thus, we hypothesized that late prenatal prescriptions might have similar consequences.
In a study from a single urban hospital, Mahon et al4 reviewed medical records for 14,786 mother/infant pairs. This study found an association between systemic erythromycin use in infants and subsequent infantile hypertrophic pyloric stenosis. A higher incidence of infantile hypertrophic pyloric stenosis in mothers using macrolides in the 10 weeks before delivery was not statistically significant. In the Mahon et al4 study, erythromycin and nonerythromycin macrolides were considered together. In contrast, in the current study, we considered the antibiotics separately because of the different effects on gastrointestinal motility, with erythromycin thought to have stronger gastrointestinal effects.11
Automated pharmacy records have been shown to be an excellent, unbiased source of prescription drug information.19–24 However, there are some limitations of these data that could cause misclassification of prescriptions. Medicaid pharmacy files only contain claims for outpatient prescriptions.25,26 Therefore, the study was not able to detect inpatient prescribing of erythromycin. Furthermore, the prescription files only indicate that a prescription was filled and cannot detect whether or not the mother took the antibiotic as prescribed. Because pharmacy data are available only for women enrolled in Medicaid or TennCare, the cohort in this study represents only 28% of the births during the study period, potentially influencing the generalizability of the findings.
The association between maternal use of nonerythromycin macrolides and pyloric stenosis was unexpected. Numbers of users of these drugs were relatively small, and the confidence interval around this estimate wide, as the findings were based on only six cases. In addition, nonerythromycin macrolide users were more likely to be urban and black than either erythromycin users or non‐users of antibiotics. An analysis of diagnostic codes associated with macrolide prescriptions indicated the presence of sexually transmitted diseases in a higher proportion of nonerythromycin macrolide than erythromycin users. Thus, it is possible that some other unmeasured factor(s) closely associated with the use of these drugs may be responsible for the observed association.
1. Honein MA, Paulozzi LJ, Himelright IM. Infantile hypertrophic pyloric stenosis after pertussis prophylaxis with erythromycin: A case review and cohort study. Lancet 1999;354:2101–5.
2. SanFilippo JA. Infantile hypertrophic pyloric stenosis related to ingestion of erythromycin estolate: A report of five cases. J Ped Surg 1976;11:177–80.
3. Stang H. Pyloric stenosis associated with erythromycin ingested through breastmilk. Minnesota Med 1986;69: 669–70,682.
4. Mahon BE, Rosenman MB, Kleiman MB. Maternal and infant use of erythromycin and other macrolide antibiotics as risk factors for infantile hypertrophic pyloric stenosis. J Pediatr 2001;139:380–4.
5. Cooper WO, Griffin MR, Hickson GB, Ray WA. Association between erythromycin exposure and development of infantile hypertrophic pyloric stenosis among Medicaid infants. Arch Pediatr Adolesc Med. In press.
6. Applegate MS, Druschel CM. The epidemiology of infantile hypertrophic pyloric stenosis in New York State: 1983–1990. Arch Pediatr Adolesc Med 1995;149:1123–9.
7. Peeters T, Matthijs G, Depoortere I, Cachet T, Hoogmartens J, Vantrappen G. Erythromycin is a motilin receptor agonist. Am J Physiol 1989;257:G470–4.
8. Jadcherla S, Klee G, Berseth C. Regulation of migrating motor complexes by motilin and pancreatic polypeptide in human infants. Pediatr Res 1997;42:365–9.
9. Ray WA, Gigante J, Mitchel EF Jr, Hickson GB. Perinatal outcomes following implementation of TennCare. JAMA 1998;279:314–6.
10. Cooper WO, Hickson GB, Gray CL, Ray WA. Changes in continuity of enrollment among high-risk children following implementation of TennCare. Arch Pediatr Adolesc Med 1999;153:1145–9.
11. Klein JO. History of macrolide use in pediatrics. Pediatr Infect Dis J 1997;16:427–31.
12. International Classification of Diseases, 9th Revision, Clinical Modification. Washington, DC: Public Health Service, US Department of Health and Human Services, 1988.
13. Current Procedural Terminology, 4th ed., revised. Chicago, IL: American Medical Association, 1998.
14. Eskenazi B, Bracken MB. Bendectin (Debendox) as a risk factor in pyloric stenosis. Am J Obstet Gynecol 1982;144:919–24.
15. Aselton P, Jick H, Cnetow SJ, Perera DR, Hutner JR, Rothman KJ. Pyloric stenosis and maternal bendectin exposure. Am J Epidemiol 1984;120:251–6.
16. Jedd MB, Melton J, Griffin MR, Kaufman B, Hoffman AD, Broughton D, et al. Factors associated with infantile hypertrophic pyloric stenosis. Am J Dis Child 1988;142:334–7.
17. Paulozzi LJ. Is Helicobacter pylori a cause of infantile hypertrophic pyloric stenosis? Med Hypoth 2000;55:119–25.
18. Curry JI, Lander TD, Stringer MD. Erythromycin as a prokinetic agent in infants and children. Aliment Pharmacol Ther 2001;15:595–603.
19. Piper JM, Ray WA, Griffin MR. Effects of Medicaid eligibility expansion on prenatal care and pregnancy outcome in Tennessee. JAMA 1990;264:2219–23.
20. Strom BL, Carson JL. Use of automated databases for pharmacoepidemiology research. In: Armenian HK, Gordis L, Levine MM, Thacker SB, eds. Epidemiologic reviews. 12th ed. Baltimore, MD: The Johns Hopkins University School of Hygiene and Public Health, 1990:87–107.
21. West SL, Savitz DA, Koch G, Strom BL, Guess HA, Hartzema A. Recall accuracy for prescription medications: Self-report compared with database information. Am J Epidemiol 1995;142:1103–10.
22. Griffin MR, Ray WA, Fought RL, Foster MA, Hays A, Schaffner W. Monitoring the safety of childhood immunizations: Methods of augmenting computerized databases for epidemiologic studies. Am J Prev Med 1988;4:5–13.
23. Leister KA, Edwards WA, Christensen DB. A comparison of patient drug regimens as viewed by the physician, pharmacist and patient. Med Care 1981;24:658–64.
24. Johnson RE, Vollmer WM. Comparing sources of drug data about the elderly. J Am Geriatr Soc 1991;39:1079–84.
25. Ray WA, Griffin MR. Use of Medicaid data for pharmacoepidemiology. Am J Epidemiol 1989;129:837–49.
26. Cooper WO, Federspiel CF, Griffin MR, Hickson GB. New use of anticonvulsant medications among children enrolled in Tennessee Medicaid. Arch Pediatr Adolesc Med 1997;151:1242–6.