The findings of this trial lacked a sufficient sample size, as indicated by the p values, to permit reliable inferences concerning the efficacy of the tested probiotics in preventing spontaneous premature births. However, all point estimates of relative risks for the non-HPD stratum, and the actual treatment analyses in particular, were less than 0.5.
Although the number of screened women was higher than estimated, the planned sample size, based on available international and national/local information, was not attained because the attrition rates were high. The trial had to be interrupted mainly because of a lack of on-going financial support. The main sources of attrition were the higher-than-expected prevalence of clinical, vaginal pH, and Nugent exclusion conditions. In addition, the intent-to-treat effect was likely diminished because a significant proportion of the randomised women exhibited conditions that required antibiotic therapy after the initiation of treatment (such as symptomatic vaginal discharge and urinary tract infection); this problem was more frequent in the placebo group and led to a discontinuation of the treatment. The intent-to-treat effect was also affected by a higher-than-expected lack of adherence, which was more frequent in the intervention group. Several randomly assigned women were excluded and discontinued treatment because they suffered abortion or developed conditions for exclusion that were strongly associated with the indication of preterm delivery, such as hypertension, and were referred to high-risk care (Figure 1). However, these women were considered in the intent-to-treat analyses.
The participants were instructed to take the treatment/placebo until approximately the 24th-26th week of gestation. By this time, at the end of the second trimester, BV is less likely to occur and commonly remits; however, the main concern was to avoid delivery at < 34 weeks, and an eventual re-infection and progression would take approximately 8 weeks to occur .
Three out of five women who spontaneously delivered prior to the 34th week of gestation exhibited initial Nugent scores of 9 or 10; these scores were present in less than 15% of the total number randomised women, which is consistent with the hypothesis that BV is a risk factor for premature birth in the study population [8, 37–39].
Four out of five spontaneous deliveries at < 34 weeks of gestation were initiated with the preterm premature rupture of membranes (PPROM); only one woman was admitted for premature labour without PPROM. These 5 cases could be associated with intrauterine infection, including BV [20, 40–43]. The non-PPROM case was not associated with any recorded maternal or foetal condition, other than BV, that can cause PD. Two cases of PPROM were associated with oligohydramnios; such an association has been shown to be related to infection. One case of PPROM occurred in a woman with HPD, and the remaining case of PPROM was not associated with any recorded maternal or foetal condition, other than BV, that could cause PPROM and PD .
Because HPD is considered to be a major predictor of subsequent PD, including very premature births , women were stratified for HPD and randomly assigned with a stratum-specific randomisation list. The number of HPD women enrolled in the trial was relatively small (35 cases); the corresponding intervention arm received more severe cases of BV than the placebo arm (Table 2). Table 3a, which shows the related outcomes, has two blank cells that indicate missing data, which prevents any meaningful analysis. Furthermore, the fact that such HPD cases are generally considered especially difficult to understand and solve [7, 42] prompted the authors to analyse the data from the non-HPD group separately.
The study protocol used ultrasound as the standard method for determining gestational age. Two premature infants had ultrasound-determined gestational ages of 30 6/7 and 32 4/7 weeks, which were different from the gestational ages estimated by the attending neonatologists (33 6/7 and 30 6/7 weeks, respectively). However, none of the premature infants were born before the 30th week, which is consistent with the relatively low rate of observed morbid events and the birth weights in the study.
Neonatal records were reviewed by two research neonatologists. All 5 infants spontaneously delivered at less than 34 weeks of gestation were discharged from the hospital. Of these five cases, only one newborn was diagnosed with any of the morbid conditions under study; that infant, whose mother belonged to the placebo group and non-HPD stratum, presented with pneumonia as well as possible respiratory distress syndrome and sepsis (Table 3). Pneumonia was diagnosed based on the presence of risk factors for infection, clinical signs, and radiographic findings. Given that radiographic findings in pneumonia can be identical to those of hyaline membrane disease , it was not possible to dismiss HMD in the immediate postpartum period in this case. Sepsis was suspected because of an immature-to-total neutrophil ratio higher than 0.2, although a blood culture was not positive, as required by the Vermont Oxford Network criteria. Other sources consider an immature-to-total neutrophil ratio higher than 0.2 to be an equally or more accurate criterion for early sepsis compared with a positive blood culture result . The infant progressed well and did not develop bronchopulmonary dysplasia.
Interestingly, the case with delivery at < 34 weeks of gestation, in which the Nugent score decreased from 9 to 2 by the end of the trial (normal vaginal mucosa), was an adherent woman with history of preterm delivery from the intervention arm who initiated treatment after the 20th week of pregnancy. It has been argued that HPD is caused by chronic intrauterine infection . One explanation is that lactobacilli that colonise the vagina of such women, and of women with repeated BV, produce an insufficient amount of lactic acid and other bactericidal substances [46, 47]. In addition, the gestational age at which the vaginal infection occurs and treatment/intervention begins may interact with the maternal and foetal immune-genetic profile/response to result in birth at less than 34 weeks of gestation [48–51]. After the 20th week of gestation, the infection may have progressed to the extent that an undesired outcome could be determined [5, 19].
To better understand the significance of the outcomes in the present trial, the early (early mid-trimester) and later (late mid-trimester) cervical immune-genetic profiles (i.e., selected cytokines and related genetic polymorphisms) of the randomised women are currently being analysed and may be helpful in dismissing or reinforcing the biological plausibility and, therefore, the prospects for a positive role of probiotics in the prevention of spontaneous premature births associated with bacterial vaginosis.
The trial was approved by an Institutional Review Board and by the National Review Board (CONEP). The trial was monitored by an independent data monitoring committee, which performed two interim analyses of data related to the main (PD) and secondary outcomes (April 2007 and July 2008), including possible adverse events; the committee received special assistance for unblinding the group allocations. The committee did not recommend the interruption of the trial. The trial was conducted according to ICH/GCP regulations and local regulations for clinical trials.
The study was registered at the NIH register platform with identifier NCT00303082.
Sponsors: The trial was supported by grants from FIOCRUZ/Brazilian Health Ministry, SAS/Brazilian Health Ministry, and the State of Rio de Janeiro Research Foundation.