Correspondence: Dr R. Lamont, Northwick Park & St Mark's NHS Trust, Watford Road, Harrow, HA1 3UJ, UK.
The association between infection and spontaneous preterm labour is now well established and thought to be responsible for preterm birth in up to 40% of cases. Preterm labour that is due to infection is refractory to the use of tocolytic agents. So the knowledge that infection may be the cause is unhelpful once a woman is admitted in spontaneous preterm labour, since by that time there may be irreversible changes in the uterine cervix, which renders futile those attempts to inhibit the process. It would be much more logical to use the association between infection and spontaneous preterm labour to identify a group of women at risk and to intervene using antibiotic prophylaxis. It is important to record, that the earlier in gestation at which abnormal genital tract colonisation is detected, the greater is the risk of an adverse outcome. For example, abnormal genital tract flora at 26–32 weeks gestation is associated with preterm birth with an odds ratio (OR) of 1.4 to 2, whereas abnormal genital tract flora at 7–16 weeks gestation carries an OR of 5 to 7.5. Intervention studies have used different antibiotics in different dosage regimes by different routes of administration to patients of differing risks at different gestational ages. Not surprisingly this has led to differing results. If intervention is to be successful, the antibiotics chosen should be active against bacterial vaginosis or bacterial vaginosis-related organisms and should be used early in pregnancy in those women with the greatest degree of abnormal genital tract flora. While there is logic in using intravaginal antibiotics to deliver a heavy antibiotic load to the vagina where heavy abnormal colonisation exists, there is also logic in considering systemic antibiotics to eradicate those organisms, which have already gained access to the decidua. It may be that the greatest chance of benefit would exist if both routes of administration were combined. Yet no study has evaluated the combination of both intravaginal and systemic antibiotics to eradicate abnormal genital tract flora for the prevention of preterm birth.
The aetiology of spontaneous preterm labour is multifactorial, but there is overwhelming evidence to implicate infection as a possible cause1 in up to 40% of cases2. When a woman is admitted in spontaneous preterm labour there may be irreversible changes in the uterine cervix, which render futile those attempts to reverse the process. It would seem sensible to consider whether it is possible to predict preterm labour by detecting abnormal genital tract flora in early pregnancy. In the first instance, it is necessary to define abnormal colonisation and a simple classification of abnormal genital tract flora in pregnancy is shown in Table 1.
Table 1. Classification of abnormal genital tract flora in pregnancy.
Group B haemolytic streptococci
Bacterial vaginosis-related organisms
After pathogens, group B haemolytic streptococci is worthwhile considering in a class of its own because of the devastating effects that this organism can have on low birthweight or preterm infants. The presence of enteropharyngeal organisms, such as Escherichia coli or Enterococcus faecalis, may sometimes be considered to be part of the normal flora, but are sometimes considered pathological. Of asymptomatic women at 31 weeks gestation, 5% will have a vagina colonised by E. coli and provided these are present in small numbers, without inflammatory cells and without other enteropharyngeal or bacterial vaginosis-related organisms, this may be considered as normal. In contrast, if E. coli is found in high numbers with other abnormal organisms in the presence of inflammatory cells in a woman with ruptured membranes or spontaneous preterm labour this should be considered as abnormal. Finally, bacterial vaginosis or bacterial vaginosis-related organisms such as anaerobes, Gardnerella vaginalis or Mycoplasma hominis may be considered indicative of abnormal genital tract flora.
The history and nomenclature of bacterial vaginosis has changed over the last half century and the name of the condition has altered according to which organism was thought to be responsible at the time. We now know that bacterial vaginosis is a polymicrobial condition associated with a number of adverse sequelae in both obstetrics and gynaecology3. The aetiology of bacterial vaginosis remains unclear and may be due to changes in endocrine status or antibiotic usage. There are, however, characteristic changes in the microbiology and the biochemistry of vaginal flora.
Normal vaginal flora is dominated by lactobacilli, which, by producing lactic acid, keep the pH of the vagina below 4.5, which discourages the growth of other organisms. In bacterial vaginosis, in association with a reduction in the quality and/or quantity of lactobacilli there is a 1,000-fold increase in the growth of other organisms. In normal flora there may be an anaerobe to aerobe ratio of 2:1 or 5:1. In bacterial vaginosis, the anaerobe to aerobe ratio is between 100:1 and 1,000:1. The range of bacteria in normal flora and bacterial vaginosis remains the same, and the diagnosis must be quantitative or semi quantitative using Gram stain, rather than qualitative using culture.
Under normal conditions of low pH, lactobacilli are able to produce hydrogen peroxide (H2O2) which is toxic to bacteria in two ways: firstly by producing toxic hydroxic radicals and secondly by combining with the heavy pool of chlorine ions in the vagina to produce chloridinium ions. Under circumstances of increased alkalinity, such as bleeding in pregnancy, sexual intercourse or vaginal douching, or under circumstances where antibiotics are used or where there is a change in endocrine status, lactobacilli at high pH are less efficient at producing H2O2 which permits the overgrowth of other organisms.
Organisms associated with bacterial vaginosis, such as Mobiluncus species and anaerobes are able to produce the keto acid, succinate, which is responsible for blunting the chemotactic response and reducing the killing ability of polymorphic nuclear leucocytes (PMNLs). This results in a vicious circle of increasing numbers of organisms without an inflammatory response. Thus, there is a large concentration of potentially pathogenic organisms with no obvious cellular host response.
The Northwick Park Study
To examine the role of abnormal flora as a predictor for preterm birth, nearly 800 women were screened in the antenatal clinic at Northwick Park Hospital, Harrow, UK using a Gram stain of vaginal secretions to diagnose bacterial vaginosis. All of the women were less than 16 weeks gestation. Approximately 80% of women were considered to have normal flora (Grade I) on Gram stain. Fourteen percent of women had grade III flora on Gram stain consistent with bacterial vaginosis. Five percent had intermediate flora on Gram stain (Grade II), which was considered to be abnormal, but not full blown bacterial vaginosis. When the outcome of pregnancy was measured by pregnancy wastage (late miscarriage together with early preterm birth) this occurred in only 3.4% of those women with Grade I flora, but in women with either Grade II or III flora the incidence of pregnancy wastage was 16% (OR = 5.35 [CI 2.73–10.5); relative risk (RR) = 3.12 [CI 2.23–4.37]; P= 0.000001)4. The study concluded that there was a significant association between abnormal genital tract colonisation in early pregnancy manifested by bacterial vaginosis detected on Gram stain of vaginal secretions and subsequent late miscarriage and preterm birth. It was considered that either alone or in conjunction with other scoring systems, this may be of additional help in identifying those women who are at risk of preterm birth. A call was made at that stage to carry out an intervention study to see whether it would be possible to reduce the incidence of late miscarriage and preterm birth by eradicating abnormal genital tract flora.
A number of other studies have confirmed that women who are found to be bacterial vaginosis-positive during pregnancy have a statistically significantly increased risk of delivering preterm over those with normal vaginal flora. The earlier in pregnancy at which this abnormal colonisation is detected, the greater is the risk of an adverse outcome. In this way, screening after 26 weeks gestation, by nature of the fact that late miscarriages and early preterm births will have been excluded, is associated with a RR of preterm birth of between 1.4 and 2. In contrast, in those studies in which abnormal genital tract flora was detected at a maximum gestation of screening of 13–16 weeks the RR of an adverse outcome was 5.5 to 6.9, respectively1 (Fig. 1).
The Use of Antibiotics in the Prevention of Preterm Birth
Antibiotics may be used prophylactically for those women at high risk of preterm birth or in women who have undergone preterm prelabour rupture of the membranes. Finally, antibiotics may be given as adjuvant therapy with tocolytics for those women who are in preterm labour.
Unfortunately treatment studies have failed to show any consensus. The diagnosis of abnormal flora has varied with some studies using culture of G. vaginalis, others using clinical criteria to diagnose bacterial vaginosis and others a Gram stain of vaginal secretions. The outcome parameters used, and measures of treatment success, together with recurrence rates have varied between studies. Antibiotic regimes have also varied in the choice of agent, the dose and duration of administration and the route, whether vaginal or systemic. Some have treated women at high risk and others low risk, and some have shown that only those women with the greatest degree of abnormal flora will respond. Many of the studies gave antibiotics at late gestations and very few considered host response and host susceptibility with respect to the production of cellular or biochemical inflammatory response. As a result, we can conclude that different studies have used different diagnostic methods with different outcome parameters or different definitions of success, to treat women of differing risks, with different susceptibilities and hence different host response, with different degrees of abnormal genital tract flora, at different gestational ages, using different antibiotics in different dosage regimes by different routes of administration and not surprisingly different results.
In a randomised controlled trial, Hauth et al.5 used oral metronidazole between 22 and 24 weeks in women at high risk of preterm birth. Unfortunately antibiotics were only shown to be of benefit in those women who had bacterial vaginosis, but the study was criticised because this subgroup of women were part of a non pre-specified post hoc analysis.
Two studies6,7 using clindamycin vaginal cream found no benefit but antibiotics were given late in pregnancy. In one study, 100% of the women were treated after 20 weeks gestation6 and in the other, 60% of women were treated after 20 weeks gestation7.
Recently, in a large study in which 2 g of metronidazole was used in women with bacterial vaginosis, no reduction was found in the incidence of preterm birth8, but the study was criticised for a number of different reasons9. Although this study was said to be on a low-risk unselected population, 85% of the women were either black or Hispanic. It should be remembered that American black women are no longer a homogeneous population. Black American women have twice the incidence of preterm birth of white American women and 50% of this difference is thought to be due to infection. However, if the American black population is further divided into black American and black African, black African women have an incidence of preterm birth similar to American white women. The Carey study8 also showed a 37% placebo response rate, which is unexplained and much higher than the usual 10% observed placebo response. Up to eight weeks could elapse from diagnosis of bacterial vaginosis to randomisation by which time the incidence of bacterial vaginosis had dropped by 25%. Nearly 50% of the women were randomised and treated after 20 weeks gestation and no women were treated before 16 weeks gestation. Although metronidazole is effective against anaerobes, it has no effect against many of the aerobic organisms or more fastidious organisms such as Mycoplasma and Ureaplasma species associated with bacterial vaginosis.
There were large numbers of exclusions from the study8, such that only 22% of those women with bacterial vaginosis were evaluated and women with symptomatic bacterial vaginosis were excluded. Of 29,625 women considered for screening, only 1,953 were assigned to treatment out of 6,540 women who had bacterial vaginosis, without other conditions, such as Trichomonas vaginalis. Among the exclusions were 999 women whose reason for exclusion was simply given as ‘other’.
In a further randomised, double-blind, placebo-controlled tricentre study, 404 women at their first antenatal clinic visit between 13 and 20 weeks gestation, who were diagnosed as having bacterial vaginosis on Gram stain of vaginal secretions, were given a 3-day course of 2% clindamycin vaginal cream or placebo10. Twenty to 24 days post treatment, women were re-evaluated and if bacterial vaginosis persisted they were given a 7-day course of the original randomised study drug. In the placebo group, 10% delivered preterm compared to only 4% in the clindamycin vaginal cream group (P < 0.02). There was also a significantly higher birthweight in the clindamycin vaginal cream group10. Some women in the study who initially had abnormal genital tract flora, when asked to return for further evaluation were found to have reverted to normal flora. Because their genital tract flora had reverted to normal, these ‘revertants’ were not included in the study, but were followed up in any case. In 24% of these women, there was an abnormal outcome of pregnancy, which was similar to those women with bacterial vaginosis treated with placebo (30%), suggesting that any adverse effect must have occurred early in the pregnancy when the genital tract flora was abnormal. Clindamycin vaginal cream was also found to have its greatest effect on those women with full blown bacterial vaginosis (Grade III Gram stain) in reducing the abnormal outcome of pregnancy (17%versus 30%; P < 0.03) compared with those women with simply abnormal flora (Grade II and Grade III Gram stain). There was also a non-statistically significant reduction in adverse outcome from 27% to 19% in the placebo and clindamycin vaginal cream groups, respectively. The study showed a 2.2-fold reduction in the incidence of preterm birth and a 3.4-fold reduction in the incidence of preterm birth when treatment was initiated before 16 weeks gestation. There was a two-fold reduction in the incidence of low birthweight and a four-fold reduction in the incidence of very low birth weight. This emphasises the importance of treating early in gestation before abnormal genital tract flora has had a chance to ascend and colonise the decidua.
The failure of late antibiotics to show any benefit is demonstrated by the ORACLE study11upon which study entry was based on the subjective assessment risk of preterm birth rather than any objective measure of risk based upon screening for infection or abnormal flora. The study was criticised for having used the wrong antibiotics in the wrong women too late in pregnancy12.
The study used a composite outcome criterion and at inception, the study was largely influenced by a worry about neonatal infection with Ureaplasma urealyticum. One of the antibiotics used, erythromycin, is not advocated in the treatment of bacterial vaginosis and is ineffective against anaerobes and Mycoplasma hominis. Co-amoxiclav, the other agent used, is not useful in bacterial vaginosis. A third group received a combination of erythromycin and co-amoxiclav. Since, one antibiotic is bactericidal and the other bacteriostatic, their actions are cancelled out. One would have to conclude from the study that Ureaplasma does not cause neonatal disease, since erythromycin was unhelpful and in any case the significance of U. urealyticum was not assessed in microbiological investigations.
We estimate that 40% of women in spontaneous preterm labour are associated with infection2. Since no screening was involved in the ORACLE study, we must conclude that 60% of women in spontaneous preterm labour in the ORACLE studies had no infection present. Only 41% of the women required tocolytic therapy. Despite this, nearly 90% were undelivered after 48 hours and 85% undelivered after 7 days. The mean gestational age at delivery was 38 completed weeks, so the conclusion must be that most of these women were not in spontaneous preterm labour of infectious aetiology and so it is a self-fulfilling prophecy that antibiotics would not have any beneficial effect.
Once the inflammatory mediators, such as cytokines, have increased in feto-maternal tissues, it may be too late or harmful to try and stop the labour. The earlier in pregnancy that abnormal genital tract flora is detected, the greater is the risk of an adverse outcome. It is logical, therefore, that antibiotics used to prevent spontaneous preterm labour and preterm birth of infectious aetiology should be used as early as possible in pregnancy.
Antibiotic treatment is likely to be successful if the endpoint of the therapy is a non-infected mother and baby rather than simply term or preterm birth. Antibiotics should be used as early as possible in pregnancy. Consideration should be given to using a combination of different routes of administration of antibiotics. There is logic in using intravaginal antibiotics, because this is where the heavy load of microorganisms exists. If, however, some of these organisms have gained access to the decidua, then intravaginal antibiotics may not be of any benefit and systemic antibiotics may be necessary to eradicate those organisms which have already ascended into the upper genital tract. The treatment is more likely to be successful if used in high-risk women, (i.e. those who are susceptible to produce an inflammatory response and those who have the greatest degree of abnormal flora).