Discussion
In this large EUROmediCAT study, we did not find evidence of a strong association between CHD and macrolide exposure, and meta-analysis of our study combined with all previous literature confirmed an overall lack of a strong association. However, we did find evidence regarding a threefold or more raised risk of AVSD specifically, significantly associated with three types of macrolide, robust across analyses with different control groups and exposure comparison groups. AVSD accounts for 2% of CHD cases in EUROCAT data, so it is not surprising this does not affect the overall CHD finding. AVSD are common in babies with Down syndrome, but none of the exposed AVSD cases had Down syndrome. The majority of negative studies regarding macrolides and CHD,14 - 26 as well as those that found an association,10 -13 did not have enough power to investigate specific subgroups of CHD. We found only one study (Crideret al. 2009)17 that has investigated the association between AVSD and erythromycin and found an elevated risk, although not statistically significant (AOR 2.2, CI: 0.8 - 6.1). This was a case-control study which obtained exposure information retrospectively by interview a considerable time after exposure, thus possibly underestimating ORs.
As elaborated by Källén et al .,11 there is compelling evidence from animal and human studies that suggests macrolides could have a link with some CHDs. At clinical concentrations, macrolides can inhibit a specific cardiac potassium current (IKr) channel, expressed by hERG (human ether a-go-go related gene). This can then lead to a prolonged QT interval, causing a type of ventricular tachycardia called torsades de pointes(TdP).4,39 In a developing rat embryo, particularly during the period before the heart is inverted (corresponding to weeks 5 - 9 of human pregnancy), TdP can result in pressure changes and misdirection of blood flow in the developing cardiovascular system, that can in turn lead to hypoxia and re-oxygenation damage resulting in septal and other vessel defects.41-44
Determining the cause of birth defects is complex, as a single birth defect may have multiple causes, just as multiple birth defects can have a single cause. However, our current AVSD finding has met many of the widely accepted Bradford Hill criteria of causation,45including the large effect size, observed pattern specific to macrolides, existence of prior hypothesis regarding CHD and a plausible biological mechanism.
In other exploratory analysis, we have found robust elevated risk of diaphragmatic hernia, orofacial clefts, syndactyly, and hydrocephalus, associated with first trimester use of erythromycin, clarithromycin, azithromycin, and clindamycin respectively. None of these associations have been previously reported. However, multiple testing may have produced some spuriously significant results, and independent confirmation is necessary.
We could not confirm previous associations of macrolides with genital anomalies,21 erythromycin with anencephaly17 or limb deficiency,17and azithromycin with orofacial clefts.16 Since our 2016 literature search, a new study found an association of urinary system defects with erythromycin (AOR 2.12, CI: 1.08 – 4.17),25 which our data and other studies did not support.12,21,46-48
Our investigation of spiramycin did not find an association with any of the CAs studied, but this was the least frequent exposure and we had limited statistical power. We found only one study from the literature that investigated spiramycin, finding no increased risk.27 Further studies are needed on the teratogenic potential of this antibiotic.
Apart from CA, other adverse outcomes related to macrolide use during pregnancy have been reported, such as miscarriage 49and fetal growth restriction in animal studies of clarithromycin and azithromycin.50-52
Safety advice about the use of macrolides during pregnancy varies across different countries. In the United Kingdom, the Medicines and Healthcare products Regulatory Agency advises alternatives to clarithromycin and azithromycin should be prescribed during pregnancy.53In Sweden, the report of Källén et al .,11 in 2005 that suggested elevated risk of CHD related to erythromycin led to warning against its use in the first trimester.53 In the USA, the previous Food and Drug Agency pregnancy classification system classified erythromycin, azithromycin and clindamycin as category B (animal studies found no risk and there are no adequately controlled human studies), and clarithromycin as category C (animal studies show adverse effects, but there are no adequately controlled human studies and benefit may warrant use despite potential risks). In light of the findings from the current study, regulatory authorities should revise the safety advice on macrolide use during the first trimester of pregnancy to reflect the potential risk of adverse outcomes. As the world considers the potential use of azithromycin - chloroquine combination in the treatment of COVID -19,8 particular attention should be given to the potential teratogenic and other negative effects of azithromycin use during the first trimester of pregnancy.