A strong body of evidence suggests that infection plays a role in the pathogenesis of preterm labour and delivery. Moreover, recent evidence indicates that intrauterine infection may also be involved in the genesis of significant neonatal and infant complications such as periventricular leukomalacia, bronchopulmonary dysplasia and cerebral palsy. This article reviews the evidence linking infection with pre-term delivery.
Evidence supporting a role for infection in the onset of labour
Three lines of evidence support a role for infection in the onset of preterm labour:
• administration of bacteria or bacterial products to animals results in either abortion or labour;
• systemic maternal infections such as pyelonephritis, pneumonia, malaria, and typhoid fever are associated with the onset of labour;
• intrauterine infection is associated with preterm labour and delivery.
Infection-induced preterm labour and delivery in animals. Considerable evidence derived from animal experimentation indicates that administration of micro-organisms or microbial products (e.g. endotoxin) to pregnant animals can induce preterm labour and delivery.1–4 In 1943, Zahl and Bjerknes5 demonstrated that injection of Shigella and Salmonella endotoxin into mice and rabbits was capable of inducing abortion. Takeda and Tsuchiya6 confirmed this observation by administering Escherichia coli endotoxin to pregnant mice and rabbits. Furthermore, immunisation of animals with an anti-endotoxin antibody ameliorates the biological effect of endotoxin.7 Animal models of ascending intrauterine infection have been developed by placing bacteria through a hysteroscope into the uterine cavity of rabbits.8 Using this model, Dombroski et al.8 and Romero et al.9 have been able to induce preterm labour and delivery. Gravett et al.10 have also developed an animal model of infection-induced preterm labour in rhesus monkeys by administering bacteria directly into the amniotic cavity or into the decidua.
Extrauterine infection and preterm labour and delivery. Systemic maternal infections such as pneumonia, pyelonephritis, malaria, and typhoid fever have been associated with preterm labour and delivery.11–22 The rate of preterm delivery associated with maternal pneumonia ranges from 15% to 48%.11–14 Although the advent of antibiotic treatment has dramatically reduced maternal mortality from this condition, it has not affected the preterm delivery rate.13,14 In the pre-antibiotic era, pyelonephritis was associated with preterm delivery. Currently, pyelonephritis is associated with preterm labour but not preterm delivery,16 probably as a result of early treatment. Similarly, typhoid fever in the pre-antibiotic era carried a 60% to 80% risk of abortion and preterm labour, but this risk decreased after the introduction of antibiotic therapy.18–20 Malaria has also been associated with a 50% rate of preterm delivery. However, chemoprophylaxis seems to protect patients from preterm delivery.21,22 Recently, periodontal infection has been reported to be a risk factor for preterm delivery (Odds ratio [OR]: 7.9; attributable risk 18.2%).23 Collectively, these data support the concept that untreated systemic maternal infection is associated with preterm labour and delivery and that treatment may decrease the rate of preterm delivery in some cases (e.g. pyelonephritis, typhoid fever) but not in others (e.g. pneumonia).
Intrauterine infection and preterm labour and delivery. Although untreated systemic maternal infections confer a high relative risk (RR) of preterm labour and delivery, these conditions are rare during pregnancy; modern treatment is effective and thus their attributable risk for preterm delivery is low. On the other hand, intrauterine infection has recently been recognised as a major factor associated with preterm labour. We will review the definition, microbiology, and clinical significance of intrauterine infection.
Definition of intrauterine infection
The gold standard for the diagnosis of an intrauterine infection is a positive microbiological culture for micro-organisms. Intrauterine infection can be classified according to the location of the micro-organisms into two broad categories: intra-amniotic and extra-amniotic infections. It is possible to obtain biological material for microbiological culture from the amniotic cavity (i.e. fluid) and the chorioamniotic space.24,25 It is not easy to culture material derived from human decidua. Therefore, in practice, most studies in patients with preterm labour and delivery have focused on microbial invasion of the amniotic cavity (defined as a positive amniotic fluid culture for micro-organisms when the fluid is retrieved by transabdominal amniocentesis).
The amniotic cavity is normally sterile and therefore the isolation of any micro-organism from the amniotic fluid constitutes evidence of microbial invasion. This condition often exists in the absence of clinical signs and symptoms of infection.26 The method of amniotic fluid collection for microbiological studies is critical. The two techniques that have been used are transabdominal amniocentesis and transcervical retrieval, either by needle puncturing of the membranes or by aspiration through an intrauterine catheter. Transcervical amniotic fluid collection is associated with a high risk of contamination with vaginal flora. Therefore, when analysing the prevalence of microbial invasion of the amniotic cavity in preterm labour, we will only consider studies in which amniotic fluid was obtained by transabdominal amniocentesis.
The term ‘clinical chorioamnionitis’ refers to the clinical syndrome associated with microbial invasion of the amniotic cavity.27 This clinical syndrome only appears in a fraction of women with microbiologically proven intra-amniotic infections. We have shown that only 12.5% of women with preterm labour and intact membranes with a positive amniotic fluid culture have clinical chorioamnionitis.28 The delay in recognising the association between intrauterine infection and preterm delivery in clinical obstetrics probably stems from the fact that most intrauterine infections during pregnancy are not associated with clinical chorioamnionitis (the relationship between clinical chorioamnionitis and preterm delivery is covered later in this article).
Culture of the chorioamniotic space provides information about intrauterine extra-amniotic infections. Such studies can be conducted by separating amnion and chorion after delivery of the placenta. Surface cultures of the membranes are of limited value for intrauterine infection because positive results may represent contamination with vaginal flora as the placenta and membranes pass through the birth canal. Studies in which amniotic fluid and chorioamniotic cultures were performed in the same patient demonstrate that micro-organisms are isolated twice as frequently from the chorioamniotic space than from the amniotic fluid (20% vs. 9%).24 These observations support the view that microbial invasion of the amniotic cavity is the result of advancing disease from the extra-amniotic to the intra-amniotic space.
Pathways of intrauterine infection
Micro-organisms may gain access to the amniotic cavity and fetus via any of the following pathways:
• ascending from the vagina and the cervix;
• haematogenous dissemination through the placenta (transplacental infection);
• retrograde seeding from the peritoneal cavity through the fallopian tubes;
• accidental introduction at the time of invasive procedures such as amniocentesis, percutaneous fetal blood sampling, chorionic villous sampling, or shunting.28–33
The most common pathway of intrauterine infection is the ascending route (Fig. 1).26 Supporting evidence includes:
• histological chorioamnionitis is more common and severe at the site of membrane rupture than in other locations, such as the placental chorionic plate or umbilical cord;28
• in virtually all cases of congenital pneumonia (stillbirths or neonatal), inflammation of the chorioamniotic membranes is present;29–31
• bacteria identified in cases of congenital infections are similar to those found in the lower genital tract;32
• in twin gestations, histological chorioamnionitis is more common in the firstborn twin and has not been demonstrated only in the second twin. As the membranes of the first twin are generally opposed to the cervix, this is taken as evidence in favour of an ascending infection.32 These observations are consistent with our findings of the microbiological state of the amniotic cavity in twin gestations: in 11 of our cases with microbial invasion of the amniotic cavity diagnosed by transabdominal amniocentesis, the presenting sac was involved in all cases. When both amniotic cavities were infected, the inoculum size was larger in the presenting sac.34 These two observations provide additional support for the hypothesis that the mechanism of infection is through an ascending pathway.
We have proposed a four-stage process leading to intrauterine infection (Fig. 1).26 The first stage consists of an overgrowth of facultative organisms or the presence of pathological organisms (i.e. Neisseria gonorrhoeae) in the vagina and/or cervix. Bacterial vaginosis may be an early manifestation of stage I. Once micro-organisms gain access to the intrauterine cavity, they reside in the decidua (stage II). A localised inflammatory reaction leads to deciduitis and further extension to chorionitis. The infection may invade the fetal vessels (choriovasculitis) or proceed through the amnion (amnionitis) into the amniotic cavity, leading to an intra-amniotic infection (stage III). Rupture of the membranes is not a prerequisite for intra-amniotic infection, as micro-organisms are capable of crossing intact membranes.35 Once in the amniotic cavity, the bacteria may gain access to the fetus by different ports of entry (stage IV). Aspiration of the infected fluid by the fetus may lead to congenital pneumonia. Otitis, conjunctivitis, and omphalitis are localised infections that occur by direct invasion of micro-organisms from infected amniotic fluid. Seeding from any of these sites to the fetal circulation results in bacteraemia and sepsis. Another possible pathway for fetal sepsis is the spread of an infection located in the decidua parietalis to the decidua basalis, and from there directly to the fetal villous circulation.
Microbiology of intrauterine infection
The most common microbial isolates from the amniotic cavity from women with preterm labour and intact membranes are Ureaplasma urealyticum, Fusobacterium species, and Mycoplasma hominis.26 Fifty per cent of patients with microbial invasion have more than one micro-organism isolated from the amniotic cavity. The inoculum size varies considerably, and in 71% of the cases more than 105 colony-forming units per millilitre (cfu/mL) are found.28 Our observations are consistent with other studies supporting a role for Fusobacterium36,37 and Mycoplasma species38 in preterm labour.
The role of Chlamydia trachomatis as an intrauterine pathogen has not been clearly elucidated. This micro-organism is an important cause of cervicitis and has been recently isolated from amniotic fluid.39,40 A case of congenital pneumonia caused by C. trachomatis suggests that this micro-organism may be capable of causing ascending intra-amniotic infection.39 The uncertainty about the role of C. trachomatis in the aetiology of microbial invasion and intrauterine in-fection may be related to difficulties in isolating the micro-organisms from amniotic fluid with standard culture techniques.40 The use of polymerase chain reaction (PCR) to detect specific sequences for this micro-organism should help resolve this question.41
Estimates of the frequency and the type of micro-organisms participating in intrauterine infections based upon standard microbiological techniques (e.g. culture) are likely to change with the introduction of more sensitive methods for microbial recovery and identification. Jalava et al.42 have recently reported the use of universal primers to conserved bacterial sequences (16 s ribosomal DNA). This approach should be able to detect most, if not all, bacterial species in amniotic fluid using PCR. The role of viruses in the aetiology of subclinical and clinical chorioamnionitis remains unexplored. However, recent studies demonstrated that viral footprints can be detected in the amniotic fluid of patients with other complications of pregnancy.43 Inasmuch as viruses can cause inflammation of other body cavities (i.e. pleuritis, pericarditis, meningitis), it would be surprising if they were not involved in the aetiology of inflammation of the chorioamniotic membranes.
Microbial invasion of the amniotic cavity and preterm delivery
Studies examining the clinical circumstances surrounding preterm delivery indicate that a third of all patients present with preterm labour and intact membranes, a third is associated with preterm premature rupture of the membranes (PROM), and the remaining third result from delivery because of maternal or fetal indications (e.g. pre-eclampsia, growth retardation).44
To examine the relationship between microbial invasion of the amniotic cavity and preterm delivery, we will review the evidence supporting an association between intrauterine infection and spontaneous preterm labour (with or without intact membranes).
Microbial invasion of the amniotic cavity in patients with preterm labour and intact membranes
Table 1 displays the results of 33 studies in which amniocentesis was performed in women with preterm labour and intact membranes.36,45–76 The mean rate of positive amniotic fluid cultures was 12.8% (379/2963). Women with positive amniotic fluid cultures generally did not have clinical evidence of infection at presentation, but they were more likely to develop clinical chorioamnionitis (37.5%[60/160] vs. 9%[27/301]), to be refractory to tocolysis (85.3%[110/129] vs. 16.3%[8/49]), and to rupture their membranes spontaneously (40%[6/15] vs. 3.8%[2/52]) than women with negative amniotic fluid cultures. The earlier the gestational age at preterm birth, the more likely that microbial invasion of the amniotic cavity was present.77
Table 1. Microbial invasion of the amniotic cavity in women with preterm labour and intact membranes *Clinical chorioamnionitis is expressed as percentage of patients with positive amniotic fluid culture.
Microbial invasion of the amniotic cavity in patients with preterm PROM
Table 2 displays the results of amniotic fluid cultures in women with preterm PROM in 18 published studies.61,62,67,68,78–91 Positive amniotic fluid cultures were detected in 32.4% (473/1462) of patients, but clinical chorioamnionitis was present in only 29.7% (49/165) of cases with microbial invasion. This is probably an underestimation of the true prevalence of microbial invasion of the amniotic cavity. Available evidence indicates that women with PROM and severely reduced volume of amniotic fluid have a higher incidence of intra-amniotic infection than those without oligohydramnios.83 As women with oligohydramnios are less likely to have an amniocentesis, the bias in these studies is to underestimate the prevalence of infection. A similar bias is that women with preterm PROM admitted in labour generally do not undergo amniocentesis; these patients have a higher rate of microbial invasion of the amniotic cavity than those admitted without labour (39.3%[24/61] vs. 25.6%[41/160], P < 0.05). Furthermore, of patients who are not in labour on admission, 75% have a positive amniotic fluid culture at the time of the onset of labour.84 Therefore, studies restricted to women not in labour provide a lower rate of microbial invasion of the amni-otic cavity than those including patients in labour.
Table 2. Microbial invasion of the amniotic cavity in women with preterm PROM as determined by amniotic fluid studies obtained by transabdominal amniocentesis
Microbial invasion of the amniotic cavity in patients presenting with acute cervical incompetence
We have determined that 51.5% of patients presenting with cervical dilatation of 2 cm or more and intact membranes between 14 and 24 weeks’ gestation have a positive amniotic fluid culture for micro-organisms.92 The outcome of patients with microbial invasion was uniformly poor as they developed subsequent complications (rupture of membranes, clinical chorioamnionitis or pregnancy loss). An infection-induced spontaneous abortion may be indistinguishable from that of an incompetent cervix.
Microbial invasion of the amniotic cavity in patients with twin gestations and preterm labour
Although multiple gestation is a known risk factor for preterm delivery, the cause of preterm delivery in this setting is poorly understood. The traditional explanation for the higher rate of preterm delivery in these pregnancies has been that uterine overdistension causes premature activation of the mechanisms responsible for the initiation of labour. It is also possible, however, that the premature cervical effacement and dilatation that occurs in multiple gestations favours an ascending infection and this, in turn, leads to preterm labour or PROM. We found that microbial invasion of the amniotic cavity occurs in 11.9% of twin gestations that present with preterm labour and deliver a preterm neonate.93 This finding is in contrast to the 21.6% observed in singleton gestations with preterm labour and delivery.28 These data suggest that intra-amniotic infection is a possible cause of preterm labour and delivery in twin gestation, but they do not support the hypothesis that intra-amniotic infection is responsible for the excessive rate of preterm delivery observed in these patients. Other investigators have confirmed that microbial invasion of the amniotic cavity occurs in patients with multiple gestations and have provided different estimates of the frequency of this condition.94,95
Microbial invasion of the amniotic cavity and fetal infection
The most advanced and serious stage of ascending intrauterine infection is fetal infection (stage IV). The overall mortality rate of neonates with congenital neonatal sepsis ranges between 25% and 90%.96–100 The wide range of these studies may represent the effect of gestational age on the likelihood of survival. One study that focused on infants born before 33 weeks’ gestation found that the mortality rate was 33% for those infected and 17% for non-infected fetuses. Carroll et al.88 have reported that fetal bacteraemia is found in 33% of fetuses with positive amniotic fluid culture and 4% of those with negative amniotic fluid culture. Therefore, subclinical fetal infection is far more common than traditionally recognised.
Histological chorioamnionitis and preterm delivery
Inflammation of the placenta and membranes is a host-response to a variety of stimuli including infection and chemical injury. Traditionally, acute inflammation of the chorioamniotic membranes has been considered an indicator of amniotic fluid infection.29–33,101–104 This view has been based upon indirect evidence. Several studies have demonstrated an association between acute inflammatory lesions of the placenta and the recovery of micro-organisms from the subchorionic plate105,106 and from the chorioamniotic space. Bacteria have been recovered from the subchorionic plate from 72% of placentas with histological evidence of chorioamnionitis.38,106,107 Furthermore, there is a strong correlation between positive amniotic fluid cultures for micro-organism and histological chorioamnionitis.108,109 Moreover, Cassel et al.24,25 have reported a strong association between positive microbial cultures from material obtained from the chorioamniotic interface and histological chorioamnionitis.
Several studies have examined the prevalence of inflammation in placentas from women delivering preterm infants. We have critically reviewed these studies elsewhere.110 Collectively, the evidence indicates that there is an association between preterm birth and the occurrence of acute chorioamnionitis.
Evidence of association between clinical infection (chorioamnionitis, endometritis and neonatal sepsis) and preterm delivery
The prevalence of endometritis is higher in women delivering preterm than in those delivering at term (preterm PROM: 18.7%[38/203] vs. term PROM: 8.4%[38/454], P < 0.001; preterm intact membranes: 13.1%[36/274] vs. term intact membranes: 6.4%[120/1881], P < 0.001). These data suggest an association between postpartum infection and preterm delivery.111 The prevalence of neonatal sepsis is 4.3 per 1000 live births in premature infants, in contrast to 0.8 per 1000 live births for term infants.112 Furthermore, the lower the birthweight, the higher the prevalence of sepsis (164/1000 for 1001–1500 g; 91/1000 for 1501–2000 g; and 23/1000 for 2001–2500 g).113
The conventional interpretation of these data is that premature newborns are more susceptible to infection. The observation that at least half of the cases of sepsis are diagnosed within 48 h after delivery, together with the high incidence of microbial invasion of the amniotic cavity in women with preterm labour and PROM, calls for a reappraisal of this traditional view. We suggest that the higher incidence of sepsis in the preterm newborn is partially attributable to the higher incidence of intrauterine infection in women with preterm labour. This interpretation is consistent with the observations of Carroll et al. using fetal blood sampling in patients with preterm PROM (see ‘Microbial infection of the amniotic cavity and fetal infection’). Furthermore, we propose that the onset of preterm labour in this subpopulation may be part of the repertoire of host defence against infection.
Is the association between infection and preterm labour/delivery causal?
In the previous section, we summarised the evidence demonstrating an association between clinical and subclinical infection and preterm labour and delivery. However, the existence of an association does not mean that infection causes preterm delivery. Indeed, it has been argued that microbial invasion of the amniotic cavity is merely the consequence of labour.114,115 Determining whether or not this relationship is causal is critical and has major clinical and therapeutic implications. This section will examine the available evidence supporting a cause-effect relationship between infection and labour.
Evaluating cause-and-effect relationships is a dif-ficult task in clinical medicine. In laboratory-based research, it is possible to expose animals or cells to a specific manoeuvre (i.e. bacteria or bacterial products) and record its effect. In contrast, a clinical investigator is rarely able to design a comparable experiment in human subjects. Specifically, it is not possible to infect normal pregnant women with micro-organisms to determine whether infection causes premature labour. Therefore, investigators are often forced to resort to alternative research methods to analyse the problem of causality (for an account of the historical developments of this intellectual challenge in infectious diseases, we refer the interested reader to the book Causation and Disease by Alfred Evans116). This section reviews the methods used by clinical epidemiologists to examine the likelihood of causality.
When studying the relationship between two events or phenomena, the first step is to determine whether or not there is an association between the two. However, the existence of a statistical relationship (i.e. a probabilistic measure of association) does not provide information about direction or sequence of events. An association between the presence of micro-organisms in the amniotic cavity and the occurrence of preterm delivery (in patients in preterm labour), for example, does not mean that micro-organisms were there before preterm labour began and that they were the cause of preterm birth. An association may be related to chance, bias, or confounding factors. Chance may be responsible for a fortuitous relationship, and replication and statistical analysis are used to exclude that type of association. Biases are systematic errors introduced into studies which may be responsible for distortion of the results. A confounding factor is an independent variable that causes change in the dependent variable and that varies systematically with the variable under study. The exclusion of the effect of biases and confounding factors is not the province of statistics, but of the architecture and design of studies.
Clinical epidemiologists use several criteria to determine whether an association is likely to be causal in nature. Table 3 displays the hierarchy and strength of different cause-and-effect criteria proposed by Sackett et al.117 These criteria are based on a classic paper by Sir Bradford Hill given as his presidential address before the Section of Occupational Medicine of the Royal Society of Medicine in 1965.118 We now will apply the epidemiological criteria displayed in Table 3 to the association of preterm labour and infection.
Table 3. Importance of individual diagnostic tests in making the causal decision Adapted from Sackett et al. A Basic Science for Clinical Medicine. Copyright 1985 David L. Sackett, R. Brian Haynes and Peter Tugwell. Published by Little, Brown & Co.
This criterion refers to the plausibility between the postulated cause and the effect (e.g. animal data, pathological data). We reviewed the evidence derived from animal experimentation in an earlier section. Collectively, this evidence indicates that intrauterine infection or the systemic administration of bacterial products can induce premature labour.
This criterion describes the precision with which the occurrence of one phenomenon predicts the occurrence of another. The ideal is a one-to-one relationship, where a cause is both necessary and sufficient. Susser119 suggested that specificity is complete when one manifestation follows from only one cause. This type of specificity is not met in the context of premature labour because preterm delivery can and does occur without any microbiological and pathological evidence of infection/inflammation. Moreover, there are limited data in humans to determine the frequency with which an intrauterine infection will lead to premature labour.
Table 4 displays the outcome of patients in whom microbial invasion of the amniotic cavity was detected in the midtrimester and who were allowed to continue their pregnancy without intervention.120,121 Most patients developed either preterm labour/delivery or PROM. These data suggest that the specificity of the relationship between infection and premature labour is incomplete. What does this mean? First, a high degree of specificity is rare in biological systems. Although the causal relationship between smoking and lung cancer is widely accepted, for example, that view could be challenged because it is not specific. Lung cancer occurs in non-smokers and, of course, smoking can cause diseases other than lung cancer, such as emphysema and chronic bronchitis. Second, the formulation of ‘the necessary and sufficient cause’ can inappropriately restrict the conceptualisation of cause. In the case of premature labour, we have provided microbiological, cytological, biochemical, immunological, and pathological data indicating that preterm labour is a syndrome and that infection is only one of its possible causes.122,123
Table 4. Pregnancy outcome in women who underwent amniocentesis in the midtrimester of pregnancy
For a relationship to be causal, the proposed causal phenomena must precede the effect. In our case, infection must precede the onset of premature labour. Although this criterion might seem obvious at first, the sequence of events is not easy to determine in clinical medicine. Many investigators have inferred that microbial invasion of the amniotic cavity causes premature labour based on the observation that micro-organisms are present in the amniotic fluid of patients admitted with preterm labour. In these studies, however, amniotic fluid was retrieved after premature labour was diagnosed. Therefore, it is entirely possible that microbial invasion is the consequence of labour rather than the converse. Indeed, we recently determined that microbial invasion of the amniotic cavity occurs in 18.8% of patients with spontaneous labour at term.124 Thus, it may be argued that microbial invasion is a phenomenon associated with labour per se rather than preterm labour. Micro-organisms gain access to the sterile amniotic cavity when cervical dilation exposes intact membranes to the normal vaginal flora or immediately after membranes rupture. Microbial invasion therefore may be the consequence of labour rather than the cause of preterm labour.
Is there evidence that infection precedes premature labour and delivery? Three sets of observations support the correct temporal relationship between infection and preterm labour and delivery.
1 Subclinical microbial invasion of the amniotic cavity in the midtrimester of pregnancy leads to either spontaneous abortion or premature delivery. Gray et al.121 reported a study in which amniotic fluid cultures for Mycoplasmas were performed in 2461 second-trimester amniocenteses. The prevalence of positive amniotic fluid cultures was 0.4% (9/2461). Of the nine patients with a positive amniotic fluid culture, one had a therapeutic abortion and eight continued their pregnancies. All eight patients had a spontaneous abortion or a preterm delivery (relative risk 8.6; 95% confidence interval [CI] 4.8, 15.4). Histological chorioamnionitis and congenital pneumonia occurred in all cases. The mean interval between a positive amniotic fluid culture and spontaneous abortion was 4 weeks. Preterm delivery occurred at 24 weeks in one case and at 30 weeks in the other. This indicates that infection is a chronic process that precedes preterm delivery.
2 In a 1993 study, we found that patients with preterm PROM with positive amniotic fluid culture for Mycoplasmas (Ureaplasma urealyticum or Mycoplasma hominis) on admission had a significantly shorter amniocentesis-to-delivery interval than those with sterile amniotic fluid.125
3 Abnormal colonisation of the lower genitourinary tract with micro-organisms is a risk factor for preterm delivery. These conditions include asymptomatic bacteriuria, bacterial vaginosis, and infection with Neisseria gonorrhoeae.52,126–136 Meta-analysis of the results of four cohort studies examining the relationship between asymptomatic bacteriuria and preterm delivery indicate that patients with this condition have a higher rate of preterm delivery than non-bacteriuric patients (typical RR = 0.51, 95% CI = 0.36, 0.69)129–133 (Table 5). Similarly, several cohort studies indicated that bacterial vaginosis is a risk factor for preterm labour.52,134–136 Because asymptomatic bacteriuria and bacterial vaginosis are present before premature labour, the sequence of events supports a causative role for these conditions in preterm labour and delivery. The precise mechanisms by which these abnormal conditions lead to premature labour have not been established. We believe that they are markers for the inability of the host to control microbial proliferation in the lower genitourinary tract. Indeed, there is evidence that patients with bacterial vaginosis are more likely to have microbial invasion of the amniotic cavity than those without that condition.136
Table 5. Meta-analysis of the incidence of preterm delivery in patients with or without bacteriuria: cohort studies LBW = low birthweight. From: Romero R, Oyarzun E, Mazor M, et al. Meta-analysis of the relationship between asymptomatic bacteriuria and preterm delivery/low birthweight. Obstetrics and Gynecololgy 1989; 73:579, reprinted with permission from the American College of Obstetricians and Gynecologists.
Consistency of the association
In contrast to the laboratory-based scientist, the clinician has limited ability for exact replication of ‘an experiment’. Indeed, variation of the study conditions is the rule in clinical medicine. Clinical epidemiologists use consistency among different studies as a substitute for replication. This criterion requires that similar results be obtained in many studies with different analytical strategies and is based on the principle that different approaches should yield similar results.
Table 1 illustrates the frequency of microbial invasion of the amniotic cavity in preterm labour. It is clear that microbial invasion of the amniotic cavity is found in most studies and that patients with microbial invasion of the amniotic cavity are more likely to deliver preterm neonates. It is noteworthy that, in several studies in which negative results were found, cultures for Mycoplasmas were not performed. These studies are likely to underestimate the frequency and consequences of microbial invasion of the amniotic cavity. In the case of bacterial vaginosis, most studies (both cohort and case control) support an association between that condition and subsequent preterm labour and delivery.
Strength of the association
The stronger the association between two phenomena, the higher the likelihood of there being a causal relationship. The strength of association is a function of the relative risk and not of the P-value. A RR of more than 2 is generally considered supportive of a strong association. Other methods for assessing the strength of an association include the correlation coefficient, the regression coefficient, or the proportion of the variance accounted for by a variable or a subset of variables. For asymptomatic bacteriuria and bacterial vaginosis, the RR is between 1.4 and 2.
The likelihood of a causal relationship is increased if a dose–response gradient can be demonstrated. If possible, the next question to be asked is whether there is a dose–response gradient between the severity of the infection and the likelihood of preterm delivery. Several lines of evidence support the existence of this gradient: (1) The median concentration of bacterial endotoxin is higher in patients in preterm labour than in patients not in labour.139 (2) The microbial inoculum is significantly greater in patients with preterm PROM admitted with preterm labour than in those admitted with preterm PROM but not in labour.85 Indeed, the proportion of patients with an inoculum size of 105 cfu/mL or greater was 41.6% in patients with preterm labour and only 15% in patients not in labour (P = 0.03).85 (3) The rate of abortion/preterm delivery after the administration of Escherichia coli bacterial endotoxin to pregnant mice exhibits a clear dose–response gradient.138
It is obviously not possible to induce intrauterine infection in normal pregnant women for experimental reasons and we therefore lack direct evidence of causality. Several trials examined the effect of antibiotic treatment in patients at risk for preterm labour (i.e. those with asymptomatic bacteriuria, bacterial vaginosis, preterm PROM) or those presenting with preterm labour.
Meta-analysis of clinical trials of antibiotic administration in patients with asymptomatic bacteriuria indicates that this therapeutic modality clearly reduces the rate of low birthweight (see Table 6).129 Several studies indicate that treatment of bacterial vaginosis can reduce the rate of preterm labour and delivery while others do not.139–146 Meta-analysis of randomised clinical trials of antibiotic administration to patients with preterm PROM (Table 7) indicates that antibiotic treatment can prolong the duration of the latency period, and reduce the rate of clinical chorioamnionitis and proven neonatal sepsis.147 Antibiotic administration to patients in preterm labour with intact membranes, by contrast, does not reduce the rate of preterm delivery and does not prolong pregnancy (Table 8).148–159
Table 6. Meta-analysis of the incidence of low birthweight (LBW) in patients with bacteriuria: randomised clinical trials From: Romero R, Oyarzun E, Mazor M, et al. Meta-analysis of the relationship between asymptomatic bacteriuria and preterm delivery/low birthweight. Obstetrics and Gynecology 1989; 73:576–582. Reprinted with permission from the American College of Obstetricians and Gynecologists.
Table 7. Results of meta-analysis of six placebo controlled trials on the impact of antimicrobial treatment on maternal and fetal outcome during expectant management of preterm premature rupture of the membranes Modified from: Mercer BM, Arheart KL. Antimicrobial therapy in expectant management of preterm premature rupture of the membranes. Lancet 1995; 346: 1271–1279.
Table 8. Summary of randomised clinical trials regarding the effect of antibiotics in women with preterm labour and intact membranes
Why were antibiotics ineffective in reducing the rate of preterm birth or prolonging pregnancy in patients with preterm labour with intact membranes? Preterm labour can be considered a syndrome defined by the presence of increased uterine contractility and cervical dilatation and caused by multiple pathological processes.122,123 Ascending infection is only one of the conditions associated with the occurrence of preterm labour. It is unlikely that antibiotic administration will be effective in altering the natural history of preterm labour in patients without an infectious process.
The data reviewed in the previous section provide strong evidence of association between intrauterine infection and preterm labour and delivery. In this section, we have presented a critical appraisal of the evidence supporting the view that this association is likely to be causal in origin. This should not, however, be taken to indicate that infection is the cause of preterm labour in all cases with microbial invasion of the amniotic cavity. We believe that, in a fraction of patients, microbial invasion is a secondary phenomenon that occurs after labour has commenced. Even if infection is secondary, however, it would be inappropriate to conclude that it is either unimportant or unrelated to the final set of events that lead to preterm labour and delivery. Future work must focus on determining why some women develop ascending infections and others do not, on methods for the rapid diagnosis of these infections, and on the role of antibiotics in reducing perinatal morbidity in women with proven intrauterine infection.