Signal testing and exploratory analysis
Overall, the proportion of cases (0.30%) exposed to macrolides in the
first trimester was not significantly different to that of non-genetic
controls (0.29%; AOR 0.99; 95%CI 0.76-1.28), or genetic controls
(0.28%; AOR 1.04; 95%CI 0.77-1.40). Similar results were also obtained
for all five specific macrolides studied (Table 3).
None of the 5 signals shown in Table 1 were confirmed (Table 3).
However, although the CHD signal was not confirmed, a pattern of
increased risk for the specific heart defect atrioventricular septal
defect (AVSD) was observed for any macrolide exposure (9 exposed cases,
AOR 2.98; 95%CI 1.48-6.01); erythromycin (4 exposed cases, AOR 3.68;
95%CI 1.28-10.61); clarithromycin (2 exposed cases, AOR 6.85; 95%CI
1.41-33.32; see foot notes in Table 4); and azithromycin (3 exposed
cases, AOR 4.50; 95%CI 1.30-15.58) (Table 4). These associations were
robust across both control and exposure comparison groups (see also
Supplementary Tables S4, S5, S7 and S8).
Since the 2016 literature review generating the signal CA, one further
study has generated new signals: urinary system with all macrolides and
respiratory system with moxifloxacin.25 These CA
subgroups were both included among non-genetic controls. There was no
evidence of any increased risk for these CA subgroups compared to
genetic controls: urinary system (AOR 0.87 95%CI 0.55-1.39);
respiratory system (AOR 0.44; 95%CI 0.10-1.89) (Supplementary Table
S3).
In the exploratory analyses of signal CA subgroups previously associated
with other antibiotics, four associations were found (Table 3):
erythromycin with diaphragmatic hernia (5 exposed cases, AOR 3.19;
95%CI 1.22-8.32); clarithromycin with orofacial clefts (8 exposed
cases, AOR 2.94; 95%CI 1.04-8.30); azithromycin with syndactyly (8
exposed cases, AOR 3.80; 95%CI 1.62-8.94); and clindamycin with
hydrocephalus (3 exposed cases, AOR 6.63; 95%CI 1.46, 30.18).