Does azithromycin given to women in labour decrease ocular bacterial infection in neonates? A double-blind, randomized trial

This was a phase III, double-blind, placebo controlled, randomized trial. The trial protocol and results of the primary endpoints have already been published [8, 9]. Briefly, 829 women attending the study health facility during labour and who had previously given informed consent were randomized to receive an oral dose of 2 g of azithromycin or placebo. Mothers and their neonates were sampled throughout the first four weeks following delivery to determine the effect of treatment on vaginal, nasopharyngeal and breast milk carriage of S. pneumoniae, S. aureus and group B streptococcus. The primary trial end-point was nasopharyngeal carriage of the study bacteria in the neonate, at day six following delivery. As part of standard care in the country, newborns received tetracycline ointment in the eyes before hospital discharge. Active and passive follow-up was conducted for eight weeks after delivery.

An ocular swab was collected from infants presenting with purulent conjunctivitis during the first eight weeks of life. Samples were collected from the lower fornix using a sterile cotton swab and kept on wet ice at the clinic. All samples were transported to the laboratory within 8 h of collection for storage at −70 °C until further processing.

DNA was extracted using the QIAamp DNA Mini Kit (QIAGEN, Germany). The presence of DNA from S. pneumoniae, S. aureus, Moraxella catarrhalis, Haemophilus influenzae, C. trachomatis and N. gonorrhoeae was assayed using the FTD SPn/Staph/MC/Hi and Vaginal Swab kits (Fast-track Diagnostics, Luxembourg) run on a RotorGene 6000 real-time PCR cycler (QIAGEN, Germany). Samples were further tested for the presence of C. trachomatis DNA using droplet-digital PCR according to a previously published method [11]. Laboratory staff was blinded to allocation arm.

Pre-defined, secondary endpoints of the PregnAnZI trial included the proportion of newborns with at least one episode of purulent conjunctivitis and ocular C. trachomatis infection within the first week of life, the first four weeks of life and during the eight weeks of the follow-up period [8]. Purulent conjunctivitis caused by bacterial pathogens other than C. trachomatis was not a stated outcome of the trial and was investigated ad hoc.

Incidence of purulent conjunctivitis and purulent conjunctivitis due to defined bacterial pathogens was compared between arms. Due to the small numbers of cases seen, only the endpoint at eight weeks follow-up was considered, rather than assessing the three individual time-points specified in the original protocol [8]. Odds ratios and exact p-values for these comparisons were calculated using the cs command in Stata. Risk factors for bacterial conjunctivitis were analysed using data from both trial arms combined and from the control arm only.

The trial was approved by the Joint Gambian Government/ Medical Research Council Ethics Committee. An independent Data Safety Monitoring Board monitored the data quality and treatment safety. Written, informed consent was obtained from all women during visits to the antenatal clinic.

A total of 829 mothers and their offspring were recruited. Baseline characteristics were similar between intervention and control arms (Table 1).

Eight-hundred-forty-three babies were delivered including 13 stillbirths. Forty infants (4.7%, 40/843) presented with purulent conjunctivitis during the eight-week follow-up period. Three infants were sampled on two separate occasions (1 in the azithromycin and 2 in the placebo arm). The median age at the onset of conjunctivitis was 4 days (IQR 2 to 8). Age of onset was missing for 6 cases of conjunctivitis. The median duration of clinical signs was 7 days (IQR 4 to 9). No difference was seen in the numbers of infants presenting with purulent conjunctivitis in the azithromycin (4.3%, 18/419) versus placebo (5.2%, 22/424) arm [OR = 0.82, 95%CI (0.44–1.54), p = 0.628] (Table 2).

S. aureus was the most common bacteria, being identified in 38% (16/42) of the swabs tested (Table 2); one sample was missing in the laboratory analysis. This finding is consistent with the results of other studies in West Africa (Togo and Nigeria) that have reported S. aureus is a leading cause of bacterial neonatal conjunctivitis in this region [12, 13]. No significant difference was seen in the incidence of S. aureus infection between trial arms with an overall incidence of 1.0% (4/419) in the azithromycin and 2.8% (12/424) in the placebo arm [OR = 0.33, 95%CI (0.11–0.98), p = 0.074].

The incidence of infection with S. pneumoniae, M. catarrhalis, H. influenzae and N. gonorrhoeae was low. Although positive results for S. pneumoniae, M. catarrhalis and H. influenzae were only obtained in the placebo arm, no statistically significant difference was seen between trial arms, possibly because of the low numbers of infections detected. No sample was positive for C. trachomatis. While there is no recent data on urogenital C. trachomatis infection in The Gambia, our results are in keeping with historical data that indicates prevalence of infection is low [14, 15]. Four swabs, all in the placebo arm, showed evidence of infection with more than one bacterium; 3 co-infected with 2 bacteria and 1 with 3 bacteria. All 3 samples positive for H. influenzae were co-infected with another bacteria (S. aureus; S. pneumoniae and M. catarrhalis; S. pneumoniae).

Overall, the incidence of purulent bacterial conjunctivitis was lower in the azithromycin arm [1.2% (5/419) versus 3.8% (16/424), OR = 0.31, 95% CI (0.12–0.82), p = 0.025)] suggesting a beneficial effect of treatment. The incidence of conjunctivitis associated with gram-positive bacteria was also lower in the intervention arm [1.0% (4/419) versus 3.3% (14/424), OR = 0.28, 95%CI (0.10–0.82), p = 0.029] (Table 2). The incidence of gram-negative bacteria was lower in the intervention arm however this difference was not statistically significant [0.2% (1/419) versus 1.4% (6/424), OR = 0.17, 95% CI (0.00–1.06), p = 0.123].

Vaginal colonization of the mother with any of the trial study bacteria (S. aureus, S. pneumoniae and group B streptococcus) prior to azithromycin administration was not a risk factor for ocular infection in the neonate (both trial arms, OR = 1.36, 95%CI 0.54–3.41, p = 0.468; placebo arm only, OR = 1.51, 95%CI 0.54–4.25, p = 0.412) (Table 3). These findings however are limited as our analysis of ocular swabs targeted more pathogens than the three study bacteria that were analysed in the vaginal swabs. Recto-vaginal swabs from mothers may also have yielded a better comparison than vaginal swabs alone.

Differences in risk of neonatal bacterial conjunctivitis were not significant among infants who received tetracycline ointment in the eye at birth compared to those who did not although odds ratios were lower for those who received ointment (both trial arms, OR = 0.79, 95%CI 0.23–2.75, p = 0.728; placebo arm only, OR = 0.52, 95%CI 0.14–1.91, p = 0.404) (Table 3). While other topical antimicrobial agents are available for use in neonates, evidence suggests that prophylaxis with tetracycline ointment does result in better outcomes in comparison to others such as povidone-iodine or erythromycin [16, 17]. Premature rupture of the membranes, which is a risk factor for conjunctivitis in the low-income setting of Malawi [18], was not a risk factor in The Gambia (both trial arms, OR = 1.12, 95%CI 0.47–2.68, p = 0.829; placebo arm only, OR = 1.59, 95%CI 0.57–4.45, p = 0.450) (Table 3).

The majority of infants in the trial received ocular tetracycline ointment after delivery, as is standard care in The Gambia. This may have lowered incidence of infection or modified the distribution of pathogens thereby impacting our endpoint. Our study is also limited by statistical power, as the trial was not designed to detect differences in the incidence of conjunctivitis or bacterial conjunctivitis between study arms. While the results showed a trend of lower incidence in the azithromycin arm for almost all bacteria assayed, significant results were obtained only when grouping all purulent bacterial conjunctivitis and conjunctivitis associated with gram-positive bacteria. The majority of samples collected from the azithromycin arm were negative for all the bacteria assayed suggesting conjunctivitis in these infants may have been chemical or viral in nature. However, our panel focused only on the major causes of bacterial conjunctivitis in neonates. It is possible that some cases could have been due to less common bacteria, yet there is no reason to believe that any other bacteria would have been more prevalent in the azithromycin arm. The limited power also extends to the risk factor analysis and may explain why tetracycline ointment or the premature rupture of membranes were not associated with conjunctivitis.

Ocular swabs were collected in the absence of transport media for the detection of C. trachomatis DNA by molecular means, as purulent conjunctivitis with C. trachomatis infection was a pre-specified outcome of the trial. This prevented us from evaluating other bacterial infection by routine culture. However, PCR is generally considered more sensitive than conventional bacteriological techniques so we do not anticipate this adversely impacted our results.