The overall incidence of preterm birth in our trial was 17.1%, which is higher than the figure reported in other populations, and which is not dissimilar to the findings of our previous, smaller study (incidence 20%; 95% CI 17%–24%) that formed the basis for the sample size calculation 
. The incidence of preterm birth was the same for the two groups and our trial provided no support for our hypothesis that this regimen of prophylactic azithromycin would reduce the incidence of preterm birth and improve outcome.
Some researchers use early preterm birth (e.g., <34 wk) as their main outcome measure as neonatal mortality is higher after early preterm than late preterm birth. We chose, as the primary outcome, overall preterm birth (<37 wk) because our previous studies had shown high rates of perinatal mortality (160/1,000) associated with late preterm birth (33–36 wk) in this population 
. In addition, morbidity is greater after late preterm than term birth, even in high income communities 
. Azithromycin was, in any case, not shown in the current study to be effective in preventing early, as well as overall, preterm birth.
As far as we are aware, our studied population of unselected pregnant women in a rural population in sub-Saharan Africa is unique in having had the gestational ages of their pregnancies confirmed by ultrasound. Gestational dating by clinical examination in later pregnancy or by the date of the last menstrual period is unreliable. Many studies in low-income countries have therefore used “low birthweight” (<2.5 kg) as a surrogate for preterm birth—but it is a poor surrogate as low birthweight babies may be either small-for-gestational age at term or preterm. We are currently studying the mortality and morbidity and developmental outcome of these babies, with known gestational age at birth.
It has been convincingly argued that the results of clinical trials should be discussed against the background of the totality of evidence from other similar studies 
. Since the publication of the Cochrane review 
that incorporated data from four studies 
, results from an additional four trials of routine antibiotic prophylaxis with preterm birth as an outcome have become available 
, including APPLe (). The largest trials, by far, are APPLe and HPTN 024. HPTN 024 was, like APPLe, performed in central Africa but relied, unlike APPLe, on menstrual dates and clinical examination rather than ultrasound for gestational age assessment 
. The eight trials took place in diverse settings (high and low income), with different types of participants (e.g., unselected women, women at high risk of preterm birth by past histories, women who were predominantly HIV positive), differing timings of treatment, and different antibiotic regimens. As well as clinical heterogeneity, there was statistical heterogeneity on analysis of the pooled data (I2
, 51%) from, overall, 6,228 pregnancies. Meta-analysis, using a random effects model showed the relative risk of preterm birth (<37 wk) with routine prophylactic antibiotics to be 1.02 (95% CI 0.86–1.22).
Randomised trials of antibiotic prophylaxis in pregnancy.
It is important to try to reconcile this finding that routine antibiotic prophylaxis does not prevent preterm birth, with the considerable observational data that associates infection with preterm labour. It is possible that different antibiotics or different antibiotic regimens with more intensive treatment schedules might impact on preterm birth rates. However, more complicated antibiotic regimens would have less appeal in resource-poor settings.
Another explanation is that ascending intrauterine infection may have been overemphasised as a primary cause of preterm birth. If factors such as psychosocial stress or heavy work, for example, are important in the premature triggering of the placental corticotropin-releasing hormone (CRH) pathway that ultimately leads to parturition 
, associated premature cervical shortening and dilatation might permit secondary ascending bacterial invasion of the uterine cavity. This has been suggested in the past 
in the context of twin pregnancy in which preterm birth is common, and early cervical dilatation does occur 
. Transvaginal ultrasound scanning has shown short cervices to be a powerful predictor of preterm birth in singleton pregnancies 
At the time of planning of the trial, it was assumed that antibiotic prophylaxis during pregnancy was unlikely to confer any harm, whether or not it conferred any benefit. The publication of the follow-up of the ORACLE trial has shown that this assumption was wrong. This report showed that children of women treated with antibiotics for preterm labour (not prophylactically) were more likely to have neuro-developmental delay 
. Our study adds further weight to the conclusion that pregnant women should not be treated with antibiotics unless for specific infections and with good evidence of likely benefit.