Eur J Clin Pharmacol (2014) 70:355–360 DOI 10.1007/s00228-013-1624-3
PHARMACOEPIDEMIOLOGY AND PRESCRIPTION
Fetal safety of erythromycin. An update of Swedish data Bengt Källén & Bengt R. Danielsson
Received: 10 October 2013 / Accepted: 2 December 2013 / Published online: 20 December 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Purpose In previous studies from the Swedish Medical Birth Register, a possible association between erythromycin therapy and an increased risk for cardiovascular defects was found. Other studies using different methodology have not verified this observation. The finding resulted in a warning for the use of erythromycin in early pregnancy, followed by a marked decline in such use. The present study was conducted to follow up on the previous observations and to find methodological explanations for the variation in results in these different published studies. Methods Data on Swedish women who gave birth during the period 1996–2011 were studied using the drug information in the Medical Birth Registry, obtained from midwife interviews conducted toward the end of the first trimester. Data on the presence of congenital malformations were ascertained from national health registers. Odds ratios were then determined using the Mantel-Haenszel methodology. Results A persistent association was seen between erythromycin use and the occurrence of cardiovascular defects, with a risk estimate of 1.70 (95 % confidence interval (CI): 1.262.39), of similar strength during the first and last 8 years of observation. We discussed the contradictory findings of the other published studies and pointed out possible methodological problems that may explain the absence of an effect in studies based on prescription registers. Conclusions Our study verified an association between early pregnancy erythromycin use and infant cardiovascular
Electronic supplementary material The online version of this article (doi:10.1007/s00228-013-1624-3) contains supplementary material, which is available to authorized users. B. Källén (*) Tornblad Institute, Lund University, Biskopsgatan 7, 223 62 Lund, Sweden e-mail:
[email protected] B. R. Danielsson National Board of Health and Welfare, Stockholm, Sweden
defects; most defects were mild. The cause of this association is unclear. Keywords Cardiovascular defects . Erythromycin . Macrolides . Pregnancy
Introduction A concern was raised in 2003 that the maternal use of erythromycin in early pregnancy was associated with an increased risk for cardiovascular defects in offspring [1]. The adjusted odds ratio was 1.91 (95 % CI 1.30-2.80). A later study expanded the observation and compared the risk with that after maternal use of phenoxymethyl penicillin [2]. The odds ratio estimate for a cardiovascular defect after erythromycin was then 1.84 (1.29-2.62) and after penicillin 0.99 (0.801.23). A follow-up study found a lower odds ratio, but the two risk estimates did not differ significantly [3]. A Medicaid study done in Tennessee found that erythromycin was not found to correlate with an increased risk for a cardiac defect [4], and a large register study from Norway also found no evidence for an association between erythromycin use and cardiovascular defects [5]. In another recent study [6], the association between erythromycin and an increased risk for cardiovascular defects was rejected based on an odds ratio of 0.95 (95 % CI 0.65-1.40). This paper updates the Swedish data on erythromycin use in early pregnancy and discusses these findings in the light of the previous literature.
Materials and methods Beginning in 1994, the maternal use of drugs in early pregnancy was prospectively recorded in the Swedish Medical Birth Register [7], which covers the whole country. This information is based on midwife interviews of pregnant
356
women, carried out when they come to their first antenatal care visit, usually before the end of the first trimester. The interviews were thus made long before the presence of a congenital malformation could be known [8]. The specific question asked was: “what drugs have you taken since you became pregnant?” The drug names were given in clear text and were centrally transformed into ATC (Anatomical, Therapeutic, and Chemical) codes. From this information, women who reported the use of erythromycin (ATC-code J01FA01) were selected. For comparison, women reporting the use of macrolides other than erythromycin were also identified. Women who gave birth during the period 1996–2011 were studied. The study concerned all births in Sweden, but did not include miscarriages or elective abortions. As supplementary information, prescriptions for erythromycin given during pregnancy were also analyzed based on information from the Swedish Register of Prescribed Drugs [9] for the years 2009-2011. This register gave the dates that the prescriptions were filled, and from these dates, delivery dates, and gestational durations were calculated the week of pregnancy when the prescription was filled. This analysis was made in order to evaluate to what extent use of a prescription register could identify exposures during the organogenetic period. The Medical Birth Register also contained information on some possible confounders: year of delivery, maternal age, parity, maternal smoking in early pregnancy, prepregnancy height and weight (from which body mass index (BMI) could be calculated), and years of unwanted childlessness as a proxy of subfertility. Diagnoses of pregnancy complications including pre-existing diabetes were also available. The presence of congenital malformations in the offspring was ascertained from three sources: the Medical Birth Register, which contains neonatal diagnoses given by a pediatrician, the Birth Defect Register (previously called the Register of Congenital Malformations), and the Hospital Discharge Register, which contains all discharge diagnoses of inpatients [10]. Linkage between registers was made using the unique identification number that every Swedish citizen has, which is widely used in society and in all medical documentation. In order to reduce the effect of variability in malformation registration, a group of “relatively severe malformations” were analyzed. Excluded from the analyses were infants who only had one or more of the following conditions: preauricular tag, tongue tie, patent ductus in preterm infants, single umbilical artery, undescended testicle, hip (sub)luxation, and nevus. This reduced the number of malformed infants in the population by 31 % (see Table 1). Risk estimates were made as Mantel-Haenszel odds ratios (OR) after adjustment for year of delivery, maternal age, parity, smoking, and BMI. Subfertility had little effect on the use of erythromycin and was not included as a co-variate. Approximate 95 % confidence intervals (95 % CI) were
Eur J Clin Pharmacol (2014) 70:355–360
estimated with Miettinen’s method. When the expected number of outcomes was below 10, a risk ratio (RR) was calculated instead as the observed overexpected ratio, with an exact 95 % CI based on Poisson distribution.
Results Among a total of 1,575,847 infants born, 2,531 had mothers who reported the use of erythromycin (1.6 per 1,000) in early pregnancy. As seen in Fig. 1, the annual number of women reporting this drug was relatively constant up to 2005, and then declined markedly. During the first 8 years of the study period, the rate was 2.7 per 1,000, and during the last 8 years it was 0.7 per 1,000. Table 1 shows the estimated odds ratios. The increased risk of any congenital defect or of a relatively severe defect did not reach statistical significance, but the risk of a cardiovascular defect did, based on 43 cases. The risk estimates were very similar for the first and last 8 years, but the latter estimate was based on only nine cases and did not reach statistical significance. No effect on other major malformations than cardiovascular defects was seen. None of the women who had used erythromycin had more than one delivery with a malformed infant. Table 2 describes the cardiovascular defects observed. The majority were of rather mild clinical importance, 10 of them were unspecified. Other macrolides than erythromycin were reported by 870 women—most had used clindamycin (n =689), azithromycin (n =73) or clarithormycin (n =71). Only 12 infants had cardiovascular defects (10 of them exposed to clindamycin), and the adjusted RR was 1.34 (95 % CI 0.76–2.73). This effect could therefore be random, but the rates of cardiovascular defects after erythromycin and after other macrolides did not differ significantly (p =0.052). Concomitant drug use was compared between women who used erythromycin and other women (Table 3), and a marked excess was noted for many drug groups: drugs for gastroesophageal reflux diagnosis (GERD), oral contraceptives, other antibiotics, non-steroid anti-inflammatory drugs (NSAID), minor analgesics, antiasthmatics, and cough medicines. NSAIDs, antihypertensives, and anticonvulsants may have teratogenic properties. If infants exposed to one or more of these drugs were removed from the analysis—198 infants (two with a cardiovascular defect)—the OR for cardiovascular defects was somewhat reduced, but remained statistically significant (OR =1.53, 95 % CI 1.12-2.03). Supplemental Table 1 shows the concomitant drugs to which infants with a cardiovascular defect were exposed. Only one infant seemed to have been exposed to a drug with a likely teratogenic capacity, clomipramine.
Eur J Clin Pharmacol (2014) 70:355–360
357
Table 1 Estimated odds ratios (OR) or risk ratios (RR) with 95 % confidence intervals (95 % CI) for congenital malformations after maternal use of erythromycin in early pregnancy (n =2,531) compared with all infants born (n =1,575,847) Malformation group
With erythromycin
Total number
OR/RR
95 % CI
Any malformation Relatively severe malformations Any cardiovascular defect 1996–2003
137 99 43 34
70,339 48,499 16,153 7,179
1.14 1.18 1.70 1.69
0.96-1.36 0.96-1.44 1.26-2.29 1.20-2.29
2004–2011 Other severe malformations than cardiovascular defects
9 56
6,661 32,346
1.71 0.92
0.78-3.25a 0.70-1.21
Adjusted for year of birth, maternal age, parity, smoking, and BMI Bold figures mark statistical significance a
RR with 95 % CI from Poisson distributions
Women who had used erythromycin together with antiasthmatic drugs and/or cough medicines showed a higher RR (1.79, 95 % CI 0.88–4.10, based on only seven exposed infants with cardiovascular defects) than women who did not report such drugs (OR=1.55, 95 % CI 1.11–2.16), but these two estimates obviously did not differ significantly. When prescriptions of erythromycin were analyzed for the years 2009–2011, 275 women had got such prescriptions filled between the date of LMP and the end of the first trimester. The distribution by pregnancy week is seen in Fig. 2. If we accept an average treatment time of 7–10 days and that the women started to use the drug on the date of filling the prescription, 74 of the 275 women had exposed their embryos during the organogenetic period—less than 30 % of all women who received a prescription from an LMP and onwards during the first trimester. The corresponding figure for those who received prescriptions from conception and
Fig. 1 Rate per 1,000 women reporting the use of erythromycin in early pregnancy, 1996–2011
onwards was about 40 %. According to the midwife interviews, 55 women during these years used erythromycin since they became pregnant.
Discussion The present study is an update of a previous communication from the Swedish Medical Birth Register [2]. It verifies the previous observation of a somewhat increased risk of a cardiovascular defect in infants exposed to erythromycin in early pregnancy. This observation differs from the results of other publications on the subject [4–6, 11, 12]. One possibility is that the primary finding was a result of multiple testing; there was no prior hypothesis, but the finding was made among a large number of different tests. The finding resulted in a warning in Sweden, and in the present study
3.0
Exposures per 1000
2.5
2.0
1.5
1.0
0.5
0.0 1995
1997
1999
2001 2003 2005 Year of birth
2007
2009
2011
358
Eur J Clin Pharmacol (2014) 70:355–360
Table 2 Specification of cardiac malformations observed after maternal use of erythromycin in early pregnancy Cardiovascular defect
Number Other congenital defect
DORV+coronary vessel malformation Tricuspidal atresia+ECD+ persistent left vena cava superior Endocardial cushion defect
1
Ventricular septum defect Atrial septum defect Ventricular and atrial septum defect Unspecified cardiac chamber defect Pulmonary valve stenosis Pulmonary artery stenosis Coarctation of aorta Pulmonary or tricuspidal valve defect Patent ductus arteriosus (term) Arteriovenous aneurysm Unspecified cardiovascular defect
1
2 13 5 1
1 with CLP and “syndrome” 1 with prominent ear
1 1 1 3 1 2 1 10
1 with hypospadias, 1 with undescended testis
CLP cleft lip/palate, DORV double outlet right ventricle, ECD endocardiac cushion defect
(Fig. 1) one can see a likely result of this: a marked decline in the reported use of erythromycin in early pregnancy. This also made odds ratio estimates performed in later years uncertain, but the risk estimate was nearly identical for the period 1996– 2003 and 2004–2011, which speaks against a randomly high peak as an explanation of the primary finding. The next possible explanation is that the result was due to confounding. The previous analysis [2] showed that women using erythromycin differed from other women in some aspects (age, parity, and smoking habits), but these factors and BMI were adjusted for in the present analysis. Concomitantly used drugs may confound the analysis if such drugs were used more often together with rather than without erythromycin, and if they had an effect of their own on the occurrence of cardiovascular defects. The present analysis made it unlikely that concomitant drug use was a significant confounder. Three such drug groups have a possible or certain effect on the risk of a cardiovascular defect: NSAIDs, anticonvulsants, and antihypertensives—if infants exposed to these drugs were removed from the analysis, the odds ratio declined somewhat, but was still statistically significant. Another problem is confounding by indication, that the underlying disease that caused the use of erythromycin had an effect of its own on the cardiovascular defect risk. The analysis of concomitantly used drugs indicated that respiratory
Table 3 Concomitant drug use among women using erythromycin Drug group
Number with erythromycin
Total number
OR/RR
95 % CI
Drugs for GERD Insulin Folic acid Antihypertensives
40 9 84 11
13,020 5,174 89,290 5,354
1.87 1.04 0.82 1.20
1.38-2.55 0.48-1.98# 0.66-1.02 0.60-2.14#
Oral contraceptives Gestagens Ovarian stimulation Thyroid hormones Other antibiotics NSAIDs Minor analgesics Drugs for migraine Anticonvulsants Neuroleptics Sedatives/hypnotics Antidepressants Antiasthmatics Cough medicines Antihistamines
22 12 6 29 387 65 276 5 13 15 11 37 157 192 166
4,113 7,407 2,799 24,143 36,775 23,447 103,206 3,788 4,267 3,991 3,315 23,319 43,533 6,224 84,203
2.39 1.02 1.00 0.84 5.80 1.36 1.33 0.69 1.79 1.74 1.77 1.06 2.01 16.7 1.09
1.44-3.62# 0.88-1.80 0.37-2.17# 0.58-1.21 5.27-6.38 1.06-1.73 1.17-1.51 0.22-1.61# 0.95-2.06# 0.97-1.27# 0.88-3.16# 0.76-1.46 1.71-2.36 14.9-18.6 0.97-1.27
Odds ratios (OR) or risk ratios (RR, #) with 95 % confidence intervals GERD gastro-esophageal reflux diagnosis, NSAIDs non-steroid antiinflammatory drugs
tract infections with coughing and asthma problems seem to be a major indication for use. This could explain the excessive use of NSAIDs, minor analgesics, antiasthmatics, other antibiotics, and cough medicines. There may be a slight risk of cardiovascular defects associated with maternal asthma and use of antiasthmatic drugs [13], but it seems weaker than the registered effect of erythromycin and can hardly explain it. Against a general teratogenic effect of infections treated with erythromycin speaks the absence of a similar phenomenon after use of phenoxymethyl penicillin or most other antibiotics [2, 14]. The explanation why an effect of erythromycin on cardiac defect risk seems to be found in Sweden, but not in other studies of the subject [4–6, 10, 11] can have different explanations. One important difference is the way in which drug use has been identified. In the Swedish studies, this was made by interviews in early pregnancy, before the pregnancy outcome was known. In one study [10] of antibacterial medication use during pregnancy on the risk of birth defects, data were collected in a case-control situation some time after the birth of the infants. This study found an association between erythromycin use and the occurrence of anencephaly (based on seven cases) and transverse limb reductions (based on nine
Eur J Clin Pharmacol (2014) 70:355–360
359
Fig. 2 Number of prescriptions for erythromycin filled during the first trimester, 2009–2011
70 60
Number of women
50 40 30 20
LMP
0
Conception
10
0
cases), and a suggested relationship with ventricular septum defects (based on 20 cases, OR=1.4, 95 % CI 0.9-2.4). This study relied on retrospectively obtained exposure information and had a high rate of non-responders (about 30 %), which makes the findings questionable. The same methodological problems exist with the Slone Epidemiology Center Birth Defects Study [11], where an OR of 1.3 (95 % CI 0.6-2.6) was found, based on only 18 exposed infants with a cardiac defect. Other studies make use of various types of prescription registers [4–6]. This results in an uncertainty of whether exposure during organogenesis has occurred or not. If not, a dilution will occur and a falsely-low odds ratio will be obtained. This problem is valid also in the Swedish data set, so the estimated risk here may be lower than the true risk at exposure during the organogenetic period. If one compares exposure rates in the Norwegian [5] study (2004-2007) with the rate in the Swedish study before the “alarm”, the former rate is 9.5 per 1,000, the latter 2.7 per 1,000—and yet, the two populations and medical care systems are rather similar. There is probably some under-registration in the Swedish data, but it can hardly explain the difference. Another explanation is that a considerable number of women who received prescriptions for erythromycin never took the drug during the organogenetic period. In some studies [6], women were regarded as exposed if they received a prescription of a macrolide from the day of their last menstrual period, and as the length of treatment is often a week or a little more, most women who filled the prescription within the first 3 weeks would not have exposed their infants during organogenesis, which would then bias the odds ratio estimate towards 1.0. In the Norwegian study [5], exposure was counted if the prescription was filled after conception, which will reduce but not completely
1
2
Cardiac organogenesis
3
4
5 6 7 8 Pregnancy week
9
10
11
12
eliminate this error. When, in the Norwegian study, exposure was restricted to days 28–56 of gestation (weeks 5–8), 611 women were counted as exposed and the OR for a cardiovascular defect was 1.62 (95 % CI 0.86–3.02) and thus closer to the estimate in this paper (1.70), which was based on a fourtimes larger number of exposed defects and was statistically significant. Swedish data showed that the dilution can be rather strong when drug ascertainment from prescription registers is used, which may result in a false-negative finding. A further concern is the power of the studies. In the latest study [6], only 325 infants were exposed to erythromycin. The number of cardiovascular defects was not stated, but if we suppose a prevalence of 1 % in the population, the expected number would be only three cases. It can be noted that the majority of the cardiovascular defects found in the present study were relatively mild, and some were unspecified. Some of these may have been overdiagnosed where the infants had normal hearts. This phenomenon would decrease the risk estimate and could increase it only if knowledge of the exposure made such unspecific diagnoses more likely. Most of them had occurred before the “alarm” and before the use of erythromycin had declined. This makes this explanation unlikely. In our previous paper [2], we suggested a teratogenic mechanism of erythromycin related to its potential to cause QT prolongation and cardiac arrythmia via hERG block. In animal studies, several QT-labeled drugs have been shown to be teratogenic as a consequence of embryonic cardiac arrythmia [15, 16]. A recent study showed that the cardiac hERG channel is expressed and is functional during cardiac organogenesis across species, including the human embryo [17]; see Supplementary Material for further elaboration on this possible mechanism.
360
To conclude, the present data seem to support the idea of an effect of maternal erythromycin use on the risk for infant cardiovascular defects, and that the absence of an effect in some other studies may have methodological explanations. If true, the mechanism of action is unclear. Acknowledgments No financial support was obtained for this study. Contributions of authors The study was planned jointly. BK collected the data, made the analysis and drafted the first manuscript. BD commented on the draft and added part of the Discussion. Ethics The study was performed within the responsibilities of the National Board of Health and Welfare, and therefore no ethical approval from outside ethical committees was needed. Conflict of interest The authors declare no conflicts of interest.
References 1. Källén BAJ, Otterblad OP (2003) Maternal drug use in early pregnancy and infant cardiovascular defect. Reprod Toxicol 17:255–261 2. Källén BAJ, Otterblad Olausson P, Danielsson BR (2005) Is erythromycin therapy teratogenic in humans? Reprod Toxicol 20:209–214 3. Källén KBM (2005) Maternal drug use in early pregnancy and risk of cardiovascular defect in infant—A follow-up study. Reprod Toxicol 26:64–65 4. Cooper WO, Harnandez-Diaz S, Arbogast PG, Dudley JA, Dyer SM, Gideon PS, Hall KS, Kaltenbach LA, Ray WA (2008) Antibiotics potentially used in response to bioterrorism and the risk of major congenital malformations. Paediatr Perinat Epidemiol 23:18–28 5. Romøren M, Lindbæck M, Nordeng H (2012) Pregnancy outcome after gestational exposure to erythromycin—a population-based register study from Norway. Br J Clin Pharmacol 74:1053–1062 6. Dinur AB, Koren G, Matok I, Wiznitzer A, Uziel E, Gorodischer R, Levy A (2013) Fetal safety of macrolides. Antimicrob Agents Chemother 57:3307–3311
Eur J Clin Pharmacol (2014) 70:355–360 7. National Board of Health and Welfare (2003) Centre for Epidemiology. The Swedish Medical Birth Register—a summary of content and quality. (http://www.socialstyrelsen.se/ Publikationer2003/2003-112-3) 8. Källén B, Otterblad Olausson P (2001) Monitoring of maternal drug use and infant congenital malformations. Does loratadine cause hypospadias? Int J Risk Saf Med 14:115–119 9. Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U, Persson I, Sundström A, Westerholm B, Rosén M (2007) The new Swedish prescribed drug register— Opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf 16:726–735 10. National Board of Health and Welfare (2004) Centre for Epidemiology. Registration of congenital malformations in Swedish health registers. (http://www.socialstyrelsen.se/ Publikationer2004/2004-112-1) 11. Crider KS, Cleves MA, Reefhuis J, Berry RJ, Hobbs CA, Hu DJ (2009) Antibacterial medication during pregnancy and risk of birth defects. National Birth Defects Prevention Study. Arch Pediatr Adolesc Med 163:978–985 12. Lin KJ, Mitchell AA, Yau W-P, Louik C, Hernández-Diaz S (2013) Safety of macrolides during pregnancy. Am J Obstet Gynecol 208(221e):e1–e8 13. Källén B, Otterblad Olausson P (2007) Use of anti-asthmatic drugs during pregnancy. 3. Congenital malformations in the infant. Eur J Clin Pharmacol 63:383–388 14. Källén B (2009) Drugs during Pregnancy. Nova Biomedical Books, New York 15. Danielsson BR, Danielsson C, Nilsson MF (2007) Embryonic cardiac arrhythmia and generation of reactive oxygen species: common teratogenic mechanism for IKr blocking drugs. Reprod Toxicol 24: 42–56 16. Sköld A-C, Wellfelt K, Danielsson BR (2001) Stage-specific skeletal and visceral defects of the Ikr-blocker almokalant: Further evidence for teratogenicity via a hypoxia-related mechanism. Teratology 64: 292–300 17. Danielsson C, Brask J, Sköld AC, Genead R, Andersson A, Andersson U, Stockling K, Pehrson R, Grinnemo KH, Salari S, Hellmold H, Danielsson B, Sylvén C, Elinder F (2013) Exploration of human, rat, and rabbit embryonic cardiomyocytes suggests Kchannel block as a common teratogenic mechanism. Cardiovasc Res 97:23–32