From Williams Obstetrics:

27. Preterm Birth

Low birthweight is the term used to define infants who are born too small, and preterm or premature birth are the terms used to define infants who are born too soon. Prior to the last century, when the expression premature birth was first used, infants delivered before term were usually referred to as "weaklings" or "congenitally debilitated babies" (Cone, 1985). Indeed, prior to 1872, infants were not even weighed at birth whether full term or premature. In 1900, Ransom wrote that in the United States "of the thousands of premature infants born ... most are quietly laid away with ... little if any effort being made for their rescue." In his first edition of this textbook, Williams (1903) wrote: "Generally speaking, premature children weighing less than 1500 g (3 lbs 3 oz) have practically no chance of life, though exceptional cases have been reported." As the 20th century progressed, there was an increasing awareness that preterm infants required special care, as evidenced by the development of incubators and intensive care nurseries. Prematurity became nationally visible as the most frequent cause of death in infancy when country-wide statistics became available with the 1949 revision of the birth certificate, which specified gestational age and birthweights.

More recently, infant mortality has become a benchmark for international comparisons of health-care systems. In this regard, the United States has ranked poorly. For example, in 1995 the United States ranked 25th in the world, well behind Japan, Singapore, Germany, and most of the Scandinavian countries (National Center for Health Statistics, 1999). Those countries with higher rates of preterm delivery have higher rates of infant mortality. Moreover, within the United States, African-Americans are disproportionately affected by preterm delivery and infant mortality. As shown in Table 27-1, more than 28,000 infants died during infancy in 1998 in the United States, and 66 percent of these were within 4 weeks of birth. Moreover, preterm birth is implicated in at least two thirds of these early infant deaths.

Sachs and associates (1995) have legitimately challenged international comparisons of infant mortality, because they found enormous regional and international differences in the way preterm births are classified.

DEFINITIONS

In 1935, the American Academy of Pediatrics defined prematurity as a live-born infant weighing 2500 g or less (Cone, 1985). These criteria were used widely until it became apparent that there were discrepancies between gestational age and birthweight because of restricted fetal growth. The World Health Organization in 1961 added gestational age as a criterion for premature infants, defined as those born at 37 weeks or less. A distinction was made between low birthweight (2500 g or less) and prematurity (37 weeks or less). Others have suggested that preterm birth be defined as those infants delivered prior to the completion of 37 weeks (American College of Obstetricians and Gynecologists, 1995).

With continued improved care of the preterm infant, other definitions have been developed. For example, the Collaborative Group on Antenatal Steroid Therapy (1981) reported that the great preponderance of mortality and serious morbidity from preterm birth is prior to 34 weeks. Moreover, low birthweight, defined as less than 2500 g, has been modified now to describe very-low birthweight, infants weighing 1500 g or less; and extremely-low birthweight, those who weigh 1000 g or less. As shown in Table 27-2, almost 90 percent of live births in the United States occur at 37 weeks or later in gestation, and progressively fewer births are recorded with decreasing gestational weeks at delivery. The disproportionate contribution of multiple gestations to preterm births is discussed in Chapter 30 (p. 780).

In many industrialized countries including the United States (Fig. 27-1), the proportion of infants born before term has increased in the past 20 years (Joseph and associates, 1998). In Canada, for instance, births at 36 weeks' gestation or less increased from 6.3 percent in 1981 to 6.8 percent in 1992. This increase in preterm births has been attributed to changes in the frequency of multiple births, increases in obstetrical intervention, improved ascertainment of early preterm births, and increased use of ultrasound for estimating gestational age (Joseph and associates, 1998). Similar factors are undoubtedly operative in the rise in preterm births that has occurred in the United States.

The importance of low birthweight, which is approximately equivalent to birth before 37 weeks, as a predictor of infant death within 28 days of birth (neonatal death), is shown in Figure 27-2. Neonatal mortality rates by state are directly proportional to the delivery of low-birthweight infants. Those states with large urban populations, or where poverty is common, have the highest incidence. In response, the United States greatly expanded the Medicaid program in 1986 with the aim of increasing prenatal and other health services for pregnant women (Dubay and colleagues, 1995). Similarly, the federal government also initiated the Healthy Start Program in 1991 for the 15 urban areas with the worst infant mortality rates.

With respect to gestational age, a fetus or infant may be preterm, term, or postterm. With respect to size, the fetus or infant may be normally grown or appropriate for gestational age, small in size or small for gestational age, or overgrown and consequently large for gestational age. In recent years, the term small for gestational age has been widely used to categorize an infant whose birthweight is usually below the 10th percentile for its gestational age. Other often-used terms have included fetal growth retardation or intrauterine growth retardation. Within the past 5 years the term restriction has largely replaced retardation, because the latter may erroneously convey mental delay rather than only the intended suboptimal fetal growth (Chap. 29, p. 745). The infant whose birthweight is above the 90th percentile has been categorized as large for gestational age, and the infant whose weight is between the 10th and 90th percentiles is designated appropriate for gestational age. Thus, an infant born before term can be small or large for gestational age and still be preterm according to chronological gestational age. Moreover, some preterm infants have also suffered growth restriction in utero. It is important to recognize that preterm birth also frequently includes infants who have suffered subnormal in utero growth. Fetal growth restriction is discussed in Chapter 29, and this chapter will focus on preterm births where the only perceived problem is delivery before full maturation. Shown in Table 29-1 are birthweight percentiles for each gestational week between 20 and 44 weeks, which may be helpful in estimating when a preterm infant's birthweight is subnormal or excessive.

IMPACT OF PRETERM BIRTH

Obstetrical approaches to preterm labor and delivery are guided in large part by expectations that the obstetrician has for survival of the premature neonate as well as the therapeutic alternatives available for management of preterm labor (Bottoms and colleagues, 1997). That some very small infants do survive when provided with prolonged, very expensive intensive care has created serious problems in decision making. The obstetrician faces the challenge of effecting delivery in such a way as to optimize the status of the fetus-infant at birth, in the event that intensive care will be applied. The neonatologist in turn must make a judgment as how best to dispense the finite resources for medical care provided by the insurance carrier, the family, governmental agencies, the hospital, and the health-care team.

Aside from survival, another important issue is the quality of life achieved by quite immature, extremely-low-birthweight infants. It is apparent that appreciable compromise, both physical and intellectual, afflicts many such children. Given these concerns, at what time in gestation should obstetrical interventions be practiced? Although it is impossible to precisely set the earliest limit for neonatal survival, certain factors inevitably have an impact on the clinical decision-making process.

Obstetrical perception of viability probably influences survival of extremely-low-birthweight infants. Amon and co-workers (1992) and Bottoms and co-workers (1997) surveyed American obstetricians to determine their clinical opinions regarding intrapartum management of the severely preterm fetus requiring delivery. Intrapartum fetal heart rate monitoring was initiated at 23, 24, and 25 weeks by 10, 45, and 65 percent of respondents, respectively. Cesarean delivery was not performed at less than 24 weeks or less than 500 g fetal weight. Almost 90 percent of respondents were willing to perform cesarean delivery for fetal distress or breech presentation at 26 weeks or 750 g fetal weight. Delivery management prior to 26 weeks, or for fetuses smaller than 750 g, was variable and individualized. Haywood and associates (1994) found that physicians significantly underestimated survival and handicap-free survival rates, particularly in those infants delivered between 23 and 29 weeks. Doron and co-workers (1998) analyzed delivery room decisions by neonatologists to provide or withhold resuscitation for 41 infants born between 23 and 26 weeks' gestation. They concluded that the delivery room is not the best place to decide on withholding life support and that this decision is best made later in the neonatal course.

It is important to emphasize that the obstetrical decision not to perform cesarean delivery or use intrapartum monitoring does not necessarily imply that the fetus is "nonviable" or "written off." For example, Kitchen and co-workers (1992) analyzed the outcomes of live-born infants weighing 500 to 999 g and found that although 50 percent of the infants survived and only 7 percent were severely disabled, these outcomes were unrelated to the use of cesarean delivery or electronic monitoring.

Perceptions of the potential for survival are inevitably confused by difficulties incurred by imprecisely known gestational age. Most survival data are based upon birthweight, which may vary appreciably between 24 and 26 weeks. For example, infants born between 24 and 26 weeks can vary in weight from 435 g to 1640 g (Alexander and associates, 1996). Perinatal mortality and morbidity decreases markedly from 24 to 26 weeks' gestation. Indeed, survival increases from approximately 20 percent at 24 weeks' gestation up to 50 percent at 25 weeks, or an increase of almost 4 percent each day. Similarly, serious perinatal morbidity also markedly decreases each day from 24 to 26 weeks' gestation. Clinically, this daily marked association with improved outcomes is of immense importance when determining obstetrical management at the lowest end of gestational ages.

The birthweight-linked survival rates for live births during 1999 at Parkland Hospital are shown in Table 27-3. Survival data for infants 500 to 1500 g are remarkably similar to rates reported by Fanaroff and colleagues (1995) for the National Institutes of Health (NIH)-sponsored Neonatal Research Network. Chances for survival increase appreciably at or above 1000-g birthweight. These data indicate that survival is possible for infants weighing 500 to 750 g. Many of these extremely-low-birthweight infants, however, were growth restricted and therefore of more advanced maturity. For example, survival of a 380-g infant has been reported, but the gestational age was confirmed to be 25^3/7 weeks (Ginsberg and associates, 1990). Clearly, expectations for neonatal survival are primarily influenced by gestational age and maturity rather than simply by birthweight.

LOWER LIMIT OF SURVIVAL. The frontier for infant survival has been progressively pushed earlier into gestation primarily as a result of continued innovations in neonatal intensive care. Copper and associates (1993) compiled gestational age-specific neonatal mortality rates for 3386 live-born infants prospectively enrolled between 1982 and 1986 in the March of Dimes Multicenter Preterm Birth Prevention Project. As shown in Figure 27-3, neonatal mortality decreased from 100 percent at 23 weeks to about 10 percent at 29 weeks, with little additional improvement through 34 weeks. Virtually identical mortality rates for 2678 live-born infants delivered in England between 1990 and 1993 were reported by Rutter (1995). The period of gestation from 23 to 25 weeks poses the greatest dilemma for both the obstetrician and pediatrician. As shown in Figure 27-3, the probability of neonatal death before 26 weeks exceeds 75 percent. Stevenson and colleagues (1998) reported very similar low-birthweight outcomes from an inborn cohort of the NICHD¹ Neonatal Research Network in 1993 and 1994. Based on the "best obstetrical assignment of gestational age," death, severe infant morbidities, or both were great before 26 weeks' gestation and almost universal before 24 weeks' gestation. Bottoms and co-workers (1999), for the NICHD Maternal-Fetal Medicine Units Network, concluded that in newborns weighing less than 1000 g, ultrasound assessment of either fetal femur length or gestational age were most predictive of neonatal mortality.

LONG-TERM OUTCOMES. Whereas few infants with birthweights below 750 g were actively treated during the 1970s, beginning in the 1980s, treatment was often practiced for infants with birthweights of at least 500 g and those born at 24 weeks or more (Hack and colleagues, 1994). The high rate of significant neonatal morbidity in these tiny infants, and the likelihood of a normal life, must be weighed against the apparent triumph of survival. As shown in Figure 27-4, among 501- to 1500-g newborns in the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network Centers, mortality and major morbidity among survivors decreased from 1988 to 1994, but both occurred in half of newborns at 501- to 750-g birthweight. Allen and co-workers (1993) described outcomes to 6 months of life for infants delivered at 22 to 25 weeks, and who were aggressively supported with intensive care. No infants survived at 22 weeks, and virtually all survivors at 23 and 24 weeks had significant brain abnormalities (Table 27-4). They concluded that whether the occasional child who is born at 23 or 24 weeks and does well justifies the considerable mortality and morbidity of the majority is a question that should be discussed by parents, health-care providers, and society. More recently, Wood and colleagues (2000) confirmed these data in a similar study from the United Kingdom.

Several groups of investigators have attempted to measure long-term outcomes in infants delivered at the very frontier of survival during the 1980s. Whyte and co-workers (1993) prospectively followed for a minimum of 2 years 321 infants born at their hospital between 23 and 26 weeks. There were no survivors among infants delivered at 23 weeks, and only 6 percent of the survivors at 24 weeks did not sustain major long-term morbidity. Approximately half of infants delivered at 25 and 26 weeks were intact at a minimum of 2 years of age. Doyle and colleagues (1994) followed infants born at 24 to 26 weeks for 5 years or longer. Even fewer infants (approximately 20 percent) were totally free of impairment with longer intervals of survival. Hack and colleagues (1994) studied the health and developmental outcomes at early school age of 68 children with birthweights below 750 g who were born from 1982 through 1986. They concluded from their unique study that children with birthweights below 750 g who survived are at "serious disadvantage in every skill required for adequate performance." Specifically, 45 percent of the survivors needed special education facilities, 21 percent had subnormal intelligence quotient (IQ) (less than 70), and many had subnormal growth and visual ability. Rutter (1995) concluded in his review of outcomes in extremely preterm infants that full resuscitation and intensive care should definitely be given at 26 weeks, probably be given at 25 weeks, possibly be given at 24 weeks, but not at 23 weeks or earlier.

Vohr and co-workers (2000) assessed neurodevelopmental and functional outcomes at 18 to 22 months corrected age of 1151 (401 to 1000 g) survivors cared for in the 12 participating centers of the NICHD¹ Neonatal Research Network and who were born in 1993 and 1994. Only half of the 1151 survivors had a normal neurodevelopmental and sensory assessment, and those at the lower birthweights had markedly worse outcomes as predicted by chronic lung disease, grades 3 and 4 intraventricular hemorrhage, and periventricular leukomalacia.

Wood and associates (2000) reported results from the EPI Cure Study Group of the United Kingdom. They performed careful mental and psychomotor evaluations at a median of 30 months in 283 surviving infants born at 25 completed weeks or less. Infant neurological survival without disability was reported in 1 of 138 infants born during the 22nd week; 11 of 241 (5 percent) during the 23rd week; 45 of 382 (12 percent) during the 24th week; and 98 of 424 (23 percent) born during the 25th week.

UPPER LIMIT OF SIGNIFICANT PREMATURITY. Not only has the frontier for neonatal survival been pushed earlier into pregnancy, but survival of larger preterm infants has become as good as that for term infants. Is there a birthweight or gestational age threshold after which attempts to delay delivery are unwarranted? In an effort to address this question, Robertson and colleagues (1992) analyzed neonatal outcomes between 1983 and 1986 from five tertiary care centers in the United States. A total of 20,680 carefully dated pregnancies without complications such as diabetes or hypertension were identified. Most thresholds—defined as the gestational week at which the incidence of complications attributable to preterm delivery became indistinguishable from term infants—were between 32 and 34 weeks. Respiratory distress syndrome, although decreasing precipitously between 33 and 34 weeks (31 to 13 percent), still developed in about 6 percent of births between 35 and 38 weeks. DePalma and co-workers (1992) found that the birthweight threshold for neonatal mortality at Parkland Hospital was 1600 g, and the threshold for neonatal morbidity due to complications of preterm delivery was approximately 1900 g. They concluded that aggressive obstetrical attempts to prevent preterm births for infants whose weights exceed 1900 g offer few apparent benefits.

ECONOMIC IMPACT OF PRETERM BIRTH. The dollar cost of the resources used to care for low-birthweight infants is one measure of the national burden of preterm birth. These dollar costs are shown in Table 27-5. A proportionately small number of births in the United States (approximately 7 percent) consume more than a third of health-care expenditures during the first year of life. On an individual basis, it is not uncommon for the smallest surviving infants to incur special care nursery costs exceeding several hundred thousand dollars (Walker and associates, 1984). Costs exceeding $1 million are possible in those hospitalized for up to a year or longer. Moreover, because of the long-term outcomes already described, additional expenditures for developmental handicaps are necessary during the remainder of childhood for many infants.

Recently, St. John and associates (2000) used a cost-analysis model to determine that 12 percent of initial neonatal care dollars in the United States were spent on infants born between 24 and 26 weeks' gestation for 1989 through 1992, and that 43 percent was spent on those born at 37 weeks' gestation or later.

CAUSES OF PRETERM BIRTH

A wide spectrum of causes and demographic factors have been implicated in the birth of preterm infants (Table 27-6).

MEDICAL AND OBSTETRICAL COMPLICATIONS

Meis and colleagues (1995b, 1998) analyzed the causes of delivery before 37 weeks in a population-based study of singleton pregnancies performed in the NICHD¹ Maternal-Fetal Medicine Units Network. Approximately 28 percent of preterm births were indicated due to preeclampsia (43 percent), fetal distress (27 percent), fetal growth restriction (10 percent), abruptio placentae (7 percent), and fetal death (7 percent). The remaining 72 percent were due to spontaneous preterm labor with or without ruptured membranes. Women with placenta previa and multiple gestations, both common associations of preterm birth, were excluded from this analysis.

LIFESTYLE FACTORS

Behaviors such as cigarette smoking, poor nutrition and poor weight gain during pregnancy, and use of drugs such as cocaine or alcohol have been reported to play important roles in the incidence and outcome of low-birthweight infants. Some of this effect is undoubtedly due to restricted fetal growth (Chap. 29, p. 745) as well as preterm birth. Hickey and colleagues (1995), however, have shown that low maternal prenatal weight gain is associated specifically with an increased risk for preterm birth.

Alcohol abuse has been linked not only to preterm birth but also to substantially increased risk of brain injury in premature infants (Holzman and co-workers, 1995). DiFronza and Lew (1995) reviewed smoking during pregnancy and reported that tobacco use was responsible for 32,000 to 61,000 low-birthweight infants each year in the United States.

Other maternal factors implicated include young maternal age (Satin and co-workers, 1994), poverty (Meis and colleagues, 1995b), short stature (Kramer and colleagues, 1995), and occupational factors (Henriksen and co-workers, 1995; Luke and associates, 1995).

Another lifestyle factor that seems intuitively important, yet has seldom been formally studied, is psychological stress in the mother. Hedegaard and associates (1993) performed a prospective follow-up study of measures of psychological stress, using questionnaires in 5872 women with singleton pregnancies. A direct association was found between psychological stress in the 30th week of pregnancy and delivery before 37 weeks' gestation. Peacock and co-workers (1995) also found an association between psychological stress and preterm birth. Copper and colleagues (1996), from the NICHD¹ Maternal-Fetal Medicine Units Network Preterm Prediction Study, reported that maternal stress was associated with spontaneous preterm birth at less than 35 weeks' gestation and after adjustment for maternal demographic and behavioral characteristics.

GENETIC FACTORS

It has been observed for many years that preterm delivery is a condition that runs in families. This observation plus the recurrent nature of preterm birth and its differing prevalence between races has led to the suggestion of a genetic cause for preterm labor. Hoffman and Ward (1999) have reviewed the possible genetic factors implicated in preterm delivery.

AMNIONIC FLUID AND CHORIOAMNIONIC INFECTION

Chorioamnionic infection caused by a variety of microorganisms has emerged as a possible explanation for many heretofore unexplained cases of ruptured membranes and/or preterm labor. Although female reproductive tract infection was associated with prematurity more than 45 years ago (Knox and Hoerner, 1995), there was renewed interest when Bobbitt and Ledger (1977) implicated subclinical amnionic fluid infection as a cause of preterm labor. For example, pathogenic bacteria have typically been recovered at transabdominal amniocentesis from approximately 20 percent of women in preterm labor without evidence of overt clinical infection and with intact fetal membranes (Cox and associates, 1996a; Watts and co-workers, 1992). Hillier and colleagues (1995) and Hauth and co-workers (1998) have reported that an appreciable number of women with spontaneous preterm birth have pathogenic organisms recovered from the chorioamnion. Indeed, the frequency of recovery of these organisms is increased in women with spontaneous onset of preterm labor with and without amnion rupture, but not increased in women delivered preterm for a medical or obstetrical complication such as maternal hypertension or hemorrhage (Hauth and co-workers, 1998). The recovery of upper genital tract pathogens is, therefore, increased in women with spontaneous labor and delivery and is inversely associated with gestational age (Fig. 27-5) and birthweight (Fig. 27-6).

PATHOGENESIS. Schwarz and co-workers (1976) suggested that term labor is initiated by activation of phospholipase A₂, which cleaves arachidonic acid from within fetal membranes, thereby making free arachidonic acid available for prostaglandin synthesis. Subsequently, Bejar and colleagues (1981) reported that many microorganisms produce phospholipase A₂, and thus potentially may initiate preterm labor. Bennett and Elder (1992) have shown that common genital tract bacteria do not themselves produce the prostaglandins. Cox and associates (1989) provided data that bacterial endotoxin (lipopolysaccharide) introduced into the amnionic fluid stimulates decidual cells to produce cytokines and prostaglandins that may initiate labor. Romero and co-workers (1987, 1988) and Cox and associates (1988a) reported that endotoxin was present in the amnionic fluid. Andrews and colleagues (1995) found a markedly higher mean amnionic fluid interleukin-6 concentration versus gestational age in women with the spontaneous onset of labor compared with women with indicated deliveries (Fig. 27-7).

It has now been established that endogenous host products secreted in response to infection are responsible for many of the effects of infection. In endotoxin shock, for example, bacterial endotoxins exert their deleterious effect through the release of endogenous cell mediators (cytokines) of the inflammatory response. Similarly, preterm parturition due to infection is thought to be initiated by secretory products resulting from monocyte (macrophage) activation (Fig. 27-8). Cytokines, including interleukin-1, tumor necrosis factor, and interleukin-6, are such secretory products implicated in preterm labor. Narahara and Johnston (1993) have suggested that platelet-activating factor, which is found in the amnionic fluid, is synergistically involved in activating the cytokine network (Fig. 27-8). Platelet-activating factor is thought to be produced in the fetal lungs and kidneys. Thus, the fetus appears to play a synergistic role in the initiation of preterm birth due to bacterial infection. Teleologically, this could be advantageous to the fetus interested in extricating itself from an infected environment.

Gravett and colleagues (1994), in a remarkable experiment with rhesus monkeys, have provided the first direct evidence that infection incites preterm labor. Group B streptococci were injected into the amnionic fluid in preterm rhesus monkeys, and concentrations of cytokines and prostaglandins shown in Figure 27-8 were serially measured. Amnionic fluid cytokine concentrations increased about 9 hours after introduction of the bacteria, followed sequentially by production of the prostaglandins E₂ and F_2a and finally, uterine contractions. As observed in humans with preterm labor due to amnionic fluid infection, there was no clinical evidence of chorioamnionitis in these rhesus monkeys until after preterm labor ensued.

Although the pathway for bacteria to enter the amnionic fluid is obvious after membrane rupture, the route of access with intact membranes is unclear. Gyr and colleagues (1994) found that Escherichia coli can permeate living chorioamnionic membranes. Thus, intact fetal membranes at the cervix are not necessarily a barrier to ascending bacterial invasion of the amnionic fluid. Alternatively, the pathway for bacterial initiation of preterm labor described in Figure 27-8 may not require colonization of the amnionic fluid. For example, Cox and co-workers (1993) found that the cytokine network of cell-mediated immunity can be activated locally in decidual tissue that lines the forebag fetal membranes.

DIAGNOSIS. The interest in a possible microbial pathogenesis of preterm labor has prompted many investigators to evaluate amniocentesis for management. Garite and colleagues (1979) successfully performed sonar-directed amniocentesis in 30 of 59 women with ruptured membranes between 28 and 35 weeks. As none had clinical infection, their purpose was to establish fetal lung maturity while evaluating Gram stain and culture of amnionic fluid. Surprisingly, the fluids from nine women (30 percent) contained bacteria; six of these women developed chorioamnionitis and two neonates developed infection. Despite these reports, it has not been shown that amniocentesis used to diagnose infection is associated with improved pregnancy outcome. Feinstein and colleagues (1986) compared 73 cases of preterm ruptured membranes managed with the aid of amniocentesis with 73 matched historical controls. There were no differences in fetal condition at delivery, incidence of neonatal infection, or perinatal mortality.

In centers using amniocentesis in the management of preterm labor, several laboratory methods have been reported helpful for the rapid detection of intraamnionic infection. Romero and colleagues (1993) evaluated the diagnostic value of amnionic fluid white blood cell count, low glucose and high interleukin-6 concentration, and Gram-stained positive bacteria in 120 women with preterm labor and intact fetal membranes. Those women with positive amnionic fluid cultures were considered infected. A negative Gram stain was the most reliable test to exclude amnionic fluid bacteria (specificity 99 percent), and a high interleukin-6 was the most sensitive test (sensitivity 82 percent) in detecting amnionic fluids containing bacteria. Andrews and colleagues (1995) also found good correlation between amnionic fluid interleukin-6 and chorioamnionic microbial colonization. Yoon and associates (1996) compared the accuracy of C-reactive protein levels or white blood cell counts in maternal blood with amnionic fluid white blood cell counts and found the latter to be superior in the confirmation of amnionic fluid infection.

Gonik and colleagues (1985) reported that oligohydramnios identified by ultrasound was linked to antepartum clinical chorioamnionitis. Vintzileos and colleagues (1986) found a similar association between oligohydramnios and bacterial colonization of amnionic fluid collected by amniocentesis. Vintzileos and associates (1985) observed that fetal infection could be predicted reliably using daily biophysical profiles; however, this was not confirmed by others (Carroll and colleagues, 1995b; DeVoe and colleagues, 1994; Gauthier and co-workers, 1992; Miller and associates, 1990).

PRETERM RUPTURED MEMBRANES. An intense inflammatory reaction at the site of prematurely ruptured membranes was noted as early as 1950, and this suggested infection. McGregor and colleagues (1987) demonstrated that in vitro exposure to bacterial proteases reduced the bursting load of fetal membranes. Thus, microorganisms given access to fetal membranes may be capable of causing membrane rupture, preterm labor, or both.

BACTERIAL VAGINOSIS

Bacterial vaginosis is a condition in which the normal, hydrogen peroxide-producing lactobacillus-predominant vaginal flora is replaced with anaerobic bacteria, Gardnerella vaginalis, Mobiluncus species, and Mycoplasma hominis (Hillier and colleagues, 1995; Nugent and co-workers, 1991). Clinical diagnostic features described by Amsel and associates (1983) include:

1. Vaginal pH greater than 4.5.

2. An amine odor when vaginal secretions are mixed with potassium hydroxide.

3. Vaginal epithelial cells heavily coated with bacilli—"clue cells."

4. A homogeneous vaginal discharge.

Bacterial vaginosis can also be diagnosed with Gram staining of vaginal secretions as reported by Nugent and co-workers (1991). Typically a Gram stain of vaginal secretion in women with bacterial vaginosis shows few white cells along with a mixed flora as compared with the normal predominance of lactobacilli.

Bacterial vaginosis has been associated with spontaneous preterm birth, preterm ruptured membranes, infection of the chorion and amnion, as well as amnionic fluid infection (Hillier and colleagues, 1995; Kurki and co-authors, 1992). Platz-Christensen and colleagues (1993) have provided some evidence that bacterial vaginosis may precipitate preterm labor by a mechanism similar to the cytokine network pathway proposed for amnionic fluid bacteria (Fig. 27-8). In contrast, Thorsen and associates (1996), in a prospective study of 3600 Danish women, found that bacterial vaginosis diagnosed before 24 weeks was not related to ruptured membranes before 37 weeks or to low birthweight. The contrasting reports may relate to the diagnosis of the clinical syndrome of bacterial vaginosis. Because of its imprecise diagnosis, it is difficult to correlate each level of Gram stain score with adverse outcomes, and specifically with spontaneous preterm birth or membrane rupture. Hauth and colleagues (2000) presented data from the large prospective NICHD¹ Maternal-Fetal Medicine Units Network observational trial that confirmed a significant increase in spontaneous preterm birth when the vaginal pH is greater than 5.0 compared with 4.7 or less, and when the Gram stain score is 9 or 10 compared with 7 or 8 or less.

TRICHOMONAS AND CANDIDA VAGINITIS

Cotch and associates (1997), after adjusting for multiple confounding factors, including bacterial vaginosis, found that women with Trichomonas vaginalis were at 30 percent increased risk of having low-birthweight infants, 30 percent increased risk of preterm birth, and a nearly doubled risk of perinatal death. Meis and co-workers (1995a) examined 2929 women at 24 and 28 weeks using 10-percent potassium hydroxide wet mount preparations and found that detection of Trichomonas or Candida had no significant association with preterm birth. Most recently, Carey and colleagues (2000a) for the NICHD¹ Maternal-Fetal Medicine Units Network reported that of 617 women with asymptomatic vaginal trichomonas, 19 percent assigned to metronidazole treatment, compared with 11 percent assigned to placebo treatment, had a preterm birth (P = .004). These authors concluded that routine screening and treatment for this condition cannot be recommended.

CHLAMYDIAL INFECTION

Although Chlamydia trachomatis is the most common sexually transmitted bacterial pathogen in the United States (Webster and colleagues, 1993), the possible influence of cervical infection with this organism on preterm birth is unclear (McGregor and French, 1991). Ryan and associates (1990) used erythromycin to treat 1323 pregnant women with positive cervical cultures for Chlamydia at enrollment for prenatal care. Pregnancy outcomes in these women were compared with 1110 similar, but untreated women. Low birthweight and ruptured membranes more than 1 hour before labor were significantly decreased with erythromycin therapy. The effects on preterm birth, however, were not specified. Andrews and colleagues (2000) from the NICHD¹ Maternal-Fetal Medicine Units Network Preterm Prediction Study reported that genitourinary chlamydial infection at 24 weeks' gestation—but not at 28 weeks—detected via a ligase chain reaction assay was associated with a twofold increase in subsequent spontaneous preterm birth. The Centers for Disease Control and Prevention guidelines for screening (1993) and treatment (1998) of chlamydial infection during pregnancy are based on the prevalence of the infection in various populations, for example, teenagers, and on the likely benefit of third-trimester screening and treatment to reduce newborn ophthalmia neonatorum or pneumonitis, rather than reduce the incidence of preterm birth (Chap. 57, p. 1493).

IDENTIFICATION OF WOMEN AT RISK FOR PRETERM BIRTH

Obstetrical approaches to preterm birth have traditionally been focused primarily on treatment interventions rather than prevention of preterm labor. The first step in prevention is early identification of women at risk for preterm birth.

RISK-SCORING SYSTEMS

A risk-scoring system devised by Papiernik and modified by Creasy and colleagues (1980) has been tested in several regions of the United States. In this system, scores of 1 through 10 are given to a variety of pregnancy factors, including socioeconomic status, reproductive history, daily habits, and current pregnancy complications. Women with scores of 10 or more are considered to be at high risk for preterm delivery.

Although Creasy and associates (1980) and Covington and co-workers (1988) reported salutary results with this risk scoring coupled with an educational prevention program, the experiences of Main and colleagues (1989) in Philadelphia using this scoring system with indigent women was less satisfactory. Similar disappointing results in indigent women were obtained by Mueller-Heubach and Guzick (1989) and Owen and associates (1990b). The Creasy program also was not successful to reduce preterm birth in the multicenter randomized trial of 2395 pregnancies managed at five centers by the Collaborative Group on Preterm Birth Prevention (1993). Hueston and colleagues (1995) reviewed all published studies and found no benefits. Subsequently, the NIH²-sponsored Maternal-Fetal Medicine Network Units study showed that risk assessment failed to identify most women who have preterm delivery (Mercer and colleagues, 1996).

Although these scoring systems have been unsuccessful in identifying pregnancies at risk for preterm labor, certain features may be more useful than others in predicting the risk of preterm delivery. Complex combinations of risk scoring systems and other screening tests have been reported to increase the prediction of spontaneous preterm birth (Crane and associates, 1999; Goldenberg and co-workers, 1998).

PRIOR PRETERM BIRTH

A history of prior preterm delivery strongly correlates with subsequent preterm labor. Table 27-7 gives the incidence of recurrent spontaneous preterm birth in over 6000 Scottish women. The risk of recurrent preterm delivery for those whose first delivery was preterm increased threefold compared with women whose first infant reached term. Strikingly, almost a third of women whose first two infants were preterm subsequently delivered preterm infants during their third pregnancies. Almost identical results were obtained in an analysis of 13,967 pregnancies in Danish women (Kristensen and co-workers, 1995). Iams and colleagues (1998a) used the NICHD¹ Maternal-Fetal Medicine Units Network Preterm Prediction Study to detail the increased risk of spontaneous preterm birth before 36 weeks' gestation in women who had had a prior preterm birth. This increased risk was substantially further increased in association with a positive (³50 ug/dL) midtrimester vaginal fetal fibronectin test or in relation to cervical shortening measured with ultrasound, especially in women with a cervix measurement at or below the 10th percentile (£25 mm) at 24 weeks' gestation.

Not only are women who deliver preterm at risk themselves for recurrence, but recent evidence suggests that this risk is also transmitted to their children. Wang and associates (1995) and Porter and colleagues (1996) have found familial aggregation of preterm birth.

CERVICAL DILATATION

Asymptomatic cervical dilatation after midpregnancy has gained attention as a risk factor for preterm delivery. Some authorities have considered such dilatation to be a normal anatomical variant, particularly in parous women. Recent studies have suggested that parity alone is not sufficient to explain cervical dilatation discovered early in the third trimester. Cook and Ellwood (1996) longitudinally studied the cervix between 18 and 30 weeks using transvaginal ultrasound in both nulliparous and parous women. Cervical length and diameter were identical in both groups of women throughout these critical weeks.

Table 27-8 gives the results of routine cervical examinations performed between 26 and 30 weeks in 185 women cared for at Parkland Hospital. Approximately a fourth of the women whose cervices were dilated 2 or 3 cm delivered prior to 34 weeks. Many of these women had experienced the same complication in earlier pregnancies. Similarly, Papiernik and colleagues (1986), in a study of cervical status before 37 weeks in 4430 women, found that precocious cervical dilatation increased the risk of preterm birth. Stubbs and colleagues (1986) performed cervical examinations in 191 women between 28 and 34 weeks and found that those with dilatation of 1 cm or more, or effacement of more than 30 percent, were at increased risk for preterm delivery. Copper and associates (1995) examined 570 women at risk for preterm birth at about 28 weeks and found that cervical condition predicted delivery before 37 weeks.

ULTRASONIC MEASUREMENT OF CERVICAL LENGTH. Iams and co-workers (1996) used transvaginal sonography to measure the length of the cervix in 2915 women at approximately 24 weeks and again at 28 weeks. As shown in Figure 27-9, the mean cervical length at 24 weeks was about 35 mm, and those women with progressively shorter cervices experienced increased rates of preterm birth. These findings were supported by investigations of Hartmann and colleagues (1999). In another study performed in the NICHD¹ Maternal-Fetal Medicine Units Network, Owen and co-workers (2000) correlated cervical length at 16 to 24 weeks' gestation with subsequent preterm birth before 35 weeks. Guzman and colleagues (2001) longitudinally evaluated 237 at-risk women with singletons between 15 and 24 weeks' gestation. At all testing periods the sensitivity and negative predictive value of cervical length to predict spontaneous preterm birth prior to 30 weeks' gestation were excellent.

Transvaginal ultrasound cervical assessment requires special expertise. Yost and colleagues (1999) caution those who perform these examinations to be wary of falsely reassuring findings due to potential anatomical and technical pitfalls. In recent small randomized trials, cervical cerclage has not been found to be beneficial in preventing delivery before 35 weeks (Althuisius and co-workers, 2000; Rust and colleagues, 2000). Rust and colleagues (2001) used ultrasonic cervical measurement criteria for cervical dilatation obtained between 16 and 24 weeks to identify and randomize 55 women to cerclage and 58 to a no cerclage intervention. The survival curve, generated with respect to gestational age at delivery, showed no significant difference between groups. However, Althuisius and colleagues (2001) randomized 35 women at high risk for cervical incompetence to either therapeutic cerclage or to bed rest. Women randomized to cerclage had significantly fewer births prior to 34 weeks and less neonatal morbidity.

Although it seems clear that pregnant women with cervical dilatation and effacement diagnosed early in the third trimester by direct cervical digital examination are at increased risk for preterm birth, it has not been established that detection appreciably improves pregnancy outcome. Buekens and colleagues (1994) randomized 2719 European women to routine cervical examinations at each prenatal visit compared with 2721 women in whom cervical examinations were not performed. Knowledge of antenatal cervical dilatation did not affect any pregnancy outcome related to preterm birth or the frequency of interventions practiced in the management of preterm labor. Importantly, cervical examinations were not related to preterm ruptured membranes. Thus, prenatal cervical examinations were neither beneficial nor harmful.

SIGNS AND SYMPTOMS

In addition to painful or painless uterine contractions, symptoms such as pelvic pressure, menstrual-like cramps, watery or bloody vaginal discharge, and pain in the low back have been empirically associated with impending preterm birth. Such symptoms are thought by some to be common in normal pregnancy, and are therefore often dismissed by patients, physicians, and nurses. The importance of these signs and symptoms has been emphasized by some investigators (Iams and associates, 1990; Kragt and Keirse, 1990). Conversely, Copper and colleagues (1990) did not find these to be meaningful in the prediction of preterm birth. Iams and colleagues (1994), in a follow-up investigation to their 1990 study, found that the signs and symptoms signaling preterm labor, including uterine contractions, only appeared within 24 hours of preterm labor. Thus, these are a late warning sign of preterm birth.

FETAL FIBRONECTIN

Fibronectin is a glycoprotein produced in 20 different molecular forms by a variety of cell types, including hepatocytes, malignant cells, fibroblasts, endothelial cells, and fetal amnion. It is present in high concentrations in maternal blood and in amnionic fluid, and is thought to have a role in intercellular adhesion in relation to implantation as well as in the maintenance of adhesion of the placenta to the decidua (Leeson and colleagues, 1996). Fetal fibronectin can be detected in cervicovaginal secretions in normal pregnancies with intact membranes at term, and appears to reflect stromal remodeling of the cervix prior to labor. Lockwood and co-workers (1991) reported that detection of fetal fibronectin in cervicovaginal secretions prior to membrane rupture may be a marker for impending preterm labor. This report has stimulated considerable interest in the use of fibronectin assays for the prediction of preterm birth. Fetal fibronectin is measured using an enzyme-linked immunosorbent assay and values exceeding 50 ng/mL are considered a positive result. Contamination of the sample by amnionic fluid and maternal blood should be avoided.

Leeson and associates (1996) and Peaceman and co-workers (1996) have found that although fibronectin-positive tests are associated with preterm birth, negative results are more consistently meaningful in predicting that preterm labor will not ensue. Cox and co-workers (1996b) found cervical dilatation to be superior to detection of fibronectin to predict preterm birth. Goldenberg and colleagues (1996a) using the NICHD¹ Network Preterm Prediction Study reported that a positive cervical or vaginal fetal fibronectin at 24 weeks' gestation was a powerful predictor of subsequent spontaneous preterm birth. Most recently, Goldenberg and co-authors (2000) reported that detection of fetal fibronectin in cervical/vaginal secretions as early as 18 to 22 weeks' gestation was predictive of preterm delivery. In a meta-analysis of 27 studies, Leitich and associates (1999) also found it to be an effective predictor of preterm delivery.

Importantly, other factors such as cervical manipulation and peripartum infection can stimulate fetal fibronectin release (Goldenberg and colleagues, 1996b; Thorp and Lukes, 1996). Similarly, Jackson and colleagues (1996) have shown that human amnion cells in vitro produce fetal fibronectin when stimulated by inflammatory products implicated in the initiation of preterm labor due to infection (Fig. 27-8).

AMBULATORY UTERINE CONTRACTION TESTING

The diagnosis of preterm labor, before it is irreversibly established, is a goal of management. To this end, uterine activity monitoring, using tocodynamometry, has received considerable interest. In 1957, Smyth described an external tocodynamometer with an innovative sensor that employed the so-called guard-ring principle. The abdominal wall is flattened by an outer ring, thus permitting the inner contraction-sensing transducer to be applied more directly to the underlying uterine wall. Several preterm birth-prevention programs incorporating such a device have been commercially available since 1985 for ambulatory uterine monitoring. The contraction sensor is belted around the abdomen and connected to a small electronic recorder worn at the waist. This recorder is used to transmit uterine activity via phone on a daily basis. Patients are educated concerning signs and symptoms of preterm labor, and their attending physicians are kept apprised of their progress. The program is expensive; for example, in Dallas in 1996 the list price was $55 to $154 per day, depending on the patient's risk status.

As shown in Figure 27-10, Katz and associates (1986) found that women who subsequently had a preterm delivery experienced increased uterine activity beginning at about 30 weeks. Subsequent widespread clinical application of home uterine contraction monitoring for the purpose of preventing preterm birth has provoked considerable controversy in the United States. Currently, the American College of Obstetricians and Gynecologists (1995) continues to take the following position: "It is not clearly demonstrated that this expensive and burdensome system can be used to actually affect the rate of preterm delivery."

Recent studies continue to show that home uterine activity monitoring is ineffective in the prevention of preterm birth. In the Collaborative Home Uterine Monitoring Study (1995), sham transducers were used in 655 women and outcomes compared with 637 women with functioning monitors. Home monitoring was ineffective in the prevention of preterm birth. Iams and colleagues (1998b) longitudinally assessed uterine activity with daily home monitoring and reported on 34,908 hours of uterine contraction data from 306 women. Contraction frequency increased with advancing gestational age but did not efficiently predict preterm birth.

Dyson and colleagues (1998), in a multicenter trial conducted at Kaiser Permanente Hospital in California, also found no benefits for home uterine activity monitoring. They randomzed 2422 women at risk for preterm birth (including 844 women with twins) to weekly contact with a nurse or to contraction monitoring at home. There were no differences in delivery before 35 weeks' gestation nor in birthweight less than 1500 g or less than 2500 g. Undesirable outcomes included a significant increase in unscheduled visits to obstetricians and significantly increased prophylactic tocolytic drug therapy in women with twins.

SALIVARY ESTRIOL

Several investigators have reported an association between increased maternal salivary estriol concentration and subsequent preterm birth (Goodwin, 1996; Heine, 1999; McGregor, 1995; and their co-workers). Goodwin (1999) reviewed the potential value of maternal salivary estriol and concluded that this test requires further evaluation. We are also of the view that use of salivary estriol to predict preterm delivery is investigational.

MANAGEMENT OF PRETERM LABOR

There are a number of factors important in directing management of the woman with possible preterm labor. Foremost is its correct identification, along with whether there is accompanying membrane rupture.

DIAGNOSIS

Early differentiation between true and false labor is difficult before there is demonstrable cervical effacement and dilatation. Uterine contractions alone can be misleading because of Braxton Hicks contractions (Chap. 14, p. 355). These contractions, described as irregular, nonrhythmical, and either painful or painless, can cause considerable confusion in the diagnosis of preterm labor. Not infrequently, women who deliver before term have uterine activity that is attributed to Braxton Hicks contractions, prompting an incorrect diagnosis of false labor.

Because uterine contractions alone may be misleading, the American Academy of Pediatrics and the American College of Obstetricians and Gynecologists (1997) has proposed the following criteria to document preterm labor between 20 and 37 weeks' gestation:

1. Contractions occurring at a frequency of four in 20 minutes or eight in 60 minutes plus progressive change in the cervix.

2. Cervical dilatation greater than 1 cm.

3. Cervical effacement of 80 percent or greater.

ANTEPARTUM MANAGEMENT

It is not apparent that making prenatal care available to more women or making more visits available has reduced preterm births (Buescher and associates, 1988; Donaldson and colleagues, 1984; Fink and co-workers, 1992; Fiscella, 1995). Despite this, women with pregnancies identified to be at risk for preterm birth, and also those who present with signs and symptoms of impending preterm delivery, have become candidates for a large number of interventions intended to improve infant outcomes. In the absence of maternal or fetal indications that warrant intentional delivery, most interventions are expected to forestall preterm birth or enhance the infants' ability to cope with the extrauterine environment. In many instances, the interventions described in the following section should be considered and not necessarily recommended for clinical practice.

PRETERM PREMATURELY RUPTURED MEMBRANES. Attempts to avoid delivery when there is preterm ruptured membranes are of two primary forms:

1. Nonintervention or expectant management, in which spontaneous labor is simply awaited.

2. Intervention that may include corticosteroids, given with or without tocolytic agents to arrest preterm labor in order that the corticosteroids have sufficient time to induce fetal maturation.

The American College of Obstetricians and Gynecologists (1998a) has recently reviewed preterm ruptured membranes.

Known risk factors for preterm rupture of the membranes include preceding preterm birth (Guinn and co-workers, 1995), occult amnionic fluid infection (see p. 696), multiple fetuses (Chap. 30, p. 778), and abruptio placentae (Major and colleagues, 1995).

NATURAL HISTORY OF PRETERM MEMBRANE RUPTURE. Cox and associates (1988b) described the pregnancy outcomes of 298 consecutive women delivered following spontaneously ruptured membranes between 24 and 34 weeks. Although this complication was identified in only 1.7 percent of pregnancies, it contributed to 20 percent of all perinatal deaths during that time period. Preterm membrane rupture was found to be associated with other obstetrical complications that affect perinatal outcome, including multifetal gestation, breech presentation, chorioamnionitis, and intrapartum fetal distress. As a consequence of these complications, cesarean delivery was done in nearly 40 percent of women. At admission, 75 percent of the women were already in labor, 5 percent were delivered for other complications, and another 10 percent were delivered following spontaneous labor within 48 hours. In only 7 percent was delivery delayed 48 hours or more after membrane rupture. This latter subgroup, however, appeared to benefit from delayed delivery, because no neonatal deaths occurred. This was in contrast to a neonatal death rate of 80 per 1000 in infants delivered within 48 hours of membrane rupture. Nelson and colleagues (1994) reported similar results. In women already in labor upon admission, about half were undelivered at 48 hours after membrane rupture but only 13 percent were undelivered at 7 days. Both of these studies attest to the inevitability of preterm labor and birth after preterm membrane rupture.

The time period from preterm ruptured membranes to delivery is inversely proportional to the gestational age when the membranes ruptured (Carroll and associates, 1995a). As shown in Figure 27-11, very few days were gained when membranes ruptured during the third compared with the second trimester.

EXPECTANT MANAGEMENT. Despite an extensive literature concerning expectant management of preterm ruptured membranes, few randomized studies have been performed. An exception is the report by Garite and colleagues (1981), who studied 160 pregnancies with preterm ruptured membranes between 28 and 34 weeks. The women were divided into two groups that included expectant management only or corticosteroids plus tocolysis with either intravenous ethanol or magnesium sulfate. The authors concluded that active interventions did not improve perinatal outcomes and may have aggravated infection-related complications. Subsequent randomized studies by Garite and associates (1987) and Nelson and colleagues (1985) failed to show benefits for tocolysis, either with or without steroid therapy. Morales and co-workers (1989) studied 165 pregnancies with preterm ruptured membranes and lecithin:sphingomyelin ratios less than two. These were randomized to four treatment groups that included (1) expectant management only, (2) corticosteroids for fetal lung maturation, (3) ampicillin only, and (4) ampicillin plus corticosteroids. They concluded that ampicillin plus corticosteroids were beneficial because of less respiratory disease. Unfortunately, neonatal survival was not ultimately affected by any intervention, nor was the length of gestation.

Alexander and colleagues (2000), in a study performed in the NICHD¹ Maternal-Fetal Medicine Units Network, found that one or two digital cervical examinations performed with expectant management of ruptured membranes between 24 and 32 weeks' gestation were associated with a significantly shorter rupture-to-delivery interval (3 days) compared with avoidance of such examinations (5 days). Importantly, however, this difference in latency from rupture-to-delivery interval did not worsen maternal or neonatal outcomes.

SECOND-TRIMESTER RUPTURED MEMBRANES. There are both maternal and infant risks to be considered when contemplating expectant management of ruptured membranes before 25 to 26 weeks. Maternal risks include the consequences of uterine infection and sepsis. Fetal risks include pulmonary hypoplasia and limb compression deformities, which have been associated with prolonged periods of oligohydramnios due to ruptured membranes (Chap. 31, p. 822). Morales and Talley (1993) expectantly managed 94 women with singleton pregnancies and ruptured membranes prior to 25 weeks and found that 41 percent ultimately survived to 1 year of age and 27 percent were neurologically normal. The average time gained during expectant management was 11 days. Similar results were reported by Farooqi and colleagues (1998) and Winn and associates (2000). In contrast, Hibbard and colleagues (1993) analyzed neonatal mortality, morbidity, and health-care costs in 44 pregnancies expectantly managed after membrane rupture at 25 weeks or less. They concluded that immediate delivery of the fetus less than 24 weeks may be the most cost-effective plan. In this report, 77 percent of the women expectantly managed developed chorioamnionitis.

The volume of amnionic fluid and fetal age at membrane rupture appears to have prognostic importance in pregnancies before 26 weeks. Hadi and associates (1994) analyzed 178 pregnancies with ruptured membranes between 20 and 25 weeks' gestation and 40 percent had oligohydramnios defined as no sonographic pockets of 2 cm or greater. Virtually all those with oligohydramnios delivered before 25 weeks, whereas 85 percent of those with adequate amnionic fluid volume delivered in the third trimester. Carroll and colleagues (1995a) observed no cases of pulmonary hypoplasia in fetuses born after membrane rupture at 24 weeks or beyond, suggesting 23 weeks or less is the threshold for lung hypoplasia.

HOSPITALIZATION. Most obstetricians hospitalize women with pregnancies complicated by preterm ruptured membranes. Concerns about the costs of lengthy hospitalizations are usually moot, because most women enter labor within a week or less of membrane rupture. Carlan and co-workers (1993) randomized 67 carefully selected pregnancies with ruptured membranes to home versus hospital management. All women were hospitalized until 72 hours after rupture; 20 women were ultimately sent home and 25 others randomized to hospitalization. No benefits were found for hospitalization and maternal hospital stay was reduced by 50 percent in those sent home (14 to 7 days). Three women sent home delivered at outside hospitals because their labor was too advanced to return to the research hospital. Importantly, these investigators emphasize that this study was too small to conclude that home management was safe.

Unfortunately, perinatal outcome in surviving infants in whom labor is delayed is not always satisfactory. In a previous study from Parkland Hospital, Hankins and associates (1984) reported that 30 percent of 176 such infants required ventilator therapy. Overall, 13 percent died in the neonatal period and another 3 percent died before age 1. Care of these infants born preterm required over 5100 newborn hospital days (almost 14 years) and cost $2.3 million (in 1984 dollars) just for bed space. Importantly, follow-up to 4 years was carried out for 105 of these infants, and neurological abnormalities of varying degrees were found in 16 percent. Similarly, Spinillo and colleagues (1995) found that preterm infants delivered 48 hours or more after ruptured membranes were at high risk for subsequent moderate to severe neurodevelopmental impairment compared with infants born after spontaneous preterm labor with intact membranes.

INTENTIONAL DELIVERY. The prospect of deliberately delivering a pregnancy preterm seems imprudent, although intentional delivery was widely practiced prior to the 1970s due to fear of infection (Reid and Christian, 1974). Another more recent rationale for intentional delivery is to avoid the uncommon fetal death that occurs in pregnancies managed expectantly after preterm rupture. The typically very short interval from membrane rupture to onset of spontaneous labor also lends support to the rationale for intentional delivery, because so little time is gained from expectant management, especially later in the third trimester (Fig. 27-11).

There have been two randomized trials of intentional delivery of pregnancies complicated by preterm ruptured membranes. Mercer and colleagues (1993) randomized 93 pregnancies with ruptured membranes between 32 and 36 weeks to delivery compared with expectant management. All had documented fetal lung maturity. Intentional delivery reduced the length of maternal hospitalization and also reduced infection rates in both mothers and neonates. Cox and Leveno (1995) similarly randomized 129 women with ruptured membranes between 30 and 34 weeks. There was one fetal death (due to sepsis) in those pregnancies managed expectantly and three neonatal deaths (two of which were due to sepsis and one due to pulmonary hypoplasia) among those intentionally delivered. Neither management approach was felt to be satisfactory.

OVERT CHORIOAMNIONITIS. Assuming that no untoward perinatal outcome occurs due to an entangled or prolapsed cord or from placental abruption, the greatest concern with prolonged membrane rupture is the risk of maternal or fetal infection. If chorioamnionitis is diagnosed, prompt efforts to effect delivery, preferably vaginally, are initiated. Unfortunately, fever is the only reliable indicator for making this diagnosis; a temperature of 38°C (100.4°F) or higher accompanying ruptured membranes implies infection. Maternal leukocytosis by itself has been found to be unreliable by most investigators, and this has also been our experience.

With chorioamnionitis, fetal and neonatal morbidity are substantively increased. In a prospective study of nearly 700 women between 26 and 34 weeks with preterm ruptured membranes, Morales (1987) reported that 13 percent developed chorioamnionitis diagnosed by oral temperatures of 38°C and no other cause for fever. Infants born to women with chorioamnionitis had a fourfold increased neonatal mortality and a threefold increase in the incidence of respiratory distress, neonatal sepsis, and intraventricular hemorrhage. Alexander and colleagues (1998) studied the effects of clinical chorioamnionitis in 1367 very-low-birthweight infants delivered at Parkland Hospital. Approximately 7 percent of the infants were exposed to chorioamnionitis and the outcomes of these infants were compared with those without overt infection. Neonatal sepsis, respiratory distress syndrome, seizures in the first 24 hours of life, intraventricular hemorrhage, and periventricular leukomalacia were all increased with chorioamnionitis. It was concluded that very-low-birthweight infants were vulnerable to neurological injury attributable to chorioamnionitis. Similarly, Yoon and colleagues (2000) found that intra-amnionic infection in preterm infants was related to cerebral palsy at 3 years of age.

ACCELERATED MATURATION OF PULMONARY FUNCTION. A variety of clinical events—some well defined and others not—have been proposed to accelerate surfactant production sufficient to protect against respiratory distress. Gluck (1979) emphasized that surfactant production is likely to be accelerated remote from term in pregnancies complicated by a number of maternal or fetal conditions or stresses. Examples included chronic renal or cardiovascular disease, long-standing hypertensive disorders due to pregnancy, heroin addiction, fetal growth restriction, placental infarction, chorioamnionitis, or preterm ruptured membranes. This view is widely held, although more recent data refute this association. For example, Owen and associates (1990a) concluded that a "stressed" pregnancy (primarily pregnancy-associated hypertension) conferred a negligible fetal survival advantage. Similarly, Hallak and Bottoms (1993) reviewed 1395 pregnancies delivered between 24 and 35 weeks and found that prematurely ruptured membranes was not associated with accelerated pulmonary maturation.

ANTIMICROBIAL THERAPY. The microbial pathogenesis for preterm ruptured membranes (Fig. 27-8) has prompted investigations of various antimicrobials to forestall delivery. Mercer and Arheart (1995) reviewed 13 randomized studies concerning the efficacy of antimicrobials compared with placebo treatment for ruptured membranes before 35 weeks. There was a combined total of 1594 pregnancies reported between 1974 and 1993. A total of 10 outcomes were subjected to meta-analysis and only three showed a possible beneficial effect of antimicrobials: fewer women developed chorioamnionitis, fewer infants developed sepsis, and pregnancy was more often prolonged 7 days in those given antimicrobials. Neonatal survival was unaffected, as was the incidence of necrotizing enterocolitis, respiratory distress, or intracranial hemorrhage.

To further address this issue, the NICHD¹ Maternal-Fetal Medicine Units Network conducted a prospective randomized trial of ampicillin or amoxicillin plus erythromycin in women with preterm rupture between 24 and 32 weeks' gestation (Mercer and colleagues, 1997). Tocolysis, corticosteroid treatment, or both were not given in this trial. There were significantly fewer newborns with respiratory distress syndrome, necrotizing enterocolitis, or a composite adverse outcome in pregnancies in which antimicrobials were given. Latency from membrane rupture to delivery was significantly longer at 7, 14, and 21 days in women given antibiotics compared with placebo. At 7 days, 50 percent of these women remained undelivered compared with only 25 percent of those assigned to placebo. These benefits were apparent whether or not the patient had a positive test for cervicovaginal group B streptococcus at the time of amnion rupture.

Prolonged antimicrobial therapy in such pregnancies may have unwanted consequences. Kyle and Turner (1996) reported superinfection with Pseudomonas aeruginosa as a result of prolonged antibiotic therapy for preterm ruptured membranes. Carroll and colleagues (1996) and Mercer and associates (1999) have also voiced concern that such therapy potentially increases the risk for selection of resistant pathogens.

DIAGNOSIS AND MANAGEMENT OF PRETERM MEMBRANE RUPTURE AT PARKLAND HOSPITAL. Pregnancy complicated by preterm rupture of the membranes is managed as follows:

1. In women with possible amnion rupture one sterile speculum examination is performed to identify fluid coming from the cervix or pooled in the vagina. Demonstration of visible fluid is indicative of ruptured membranes and is usually accompanied by ultrasound examination to confirm oligohydramnios, to identify the presenting part, and to estimate gestational age. Nitrazine paper testing of vaginal pH has an appreciable false-positive rate associated with blood contamination, semen, or bacterial vaginosis. The microscopic inspection of cervicovaginal dried secretions for NaCl crystallization (ferning) also has an appreciable false-positive rate. Infrequently, a compelling maternal history of amnion rupture is not supported by either gross visualization or vaginal pooling of amnionic fluid or ultrasound confirmation of decreased amnionic fluid. In these instances, instillation of indigo carmine dye (one ampule) into the amnionic fluid and assessing leakage of blue fluid into the vagina can be helpful in ascertaining the integrity of the membranes.

Attempts are made to visualize the extent of cervical effacement and dilatation, but a digital examination is not performed.

2. If the gestational age is less than 34 weeks and there are no other maternal or fetal indications for delivery, the woman is observed closely in Labor and Delivery. Continuous fetal heart rate monitoring is employed to look for evidence of cord compression, especially if labor supervenes.

3. If the fetal heart rate is reassuring, and if labor does not follow, the woman is transferred to the High Risk Pregnancy Unit for close observation for signs of labor, infection, or fetal jeopardy.

4. If the gestational age is greater than 34 completed weeks and if labor has not begun following adequate evaluation, labor is induced with intravenous oxytocin unless contraindicated. If induction fails, cesarean delivery is performed.

5. Dexamethasone, 5 mg intramuscularly every 12 hours for 4 doses, is given for enhancement of fetal maturation.

6. When labor is diagnosed, ampicillin, 2 g, is given intravenously every 6 hours prior to delivery for prevention of group B streptococcal infection in the neonate.

PRETERM LABOR WITH INTACT FETAL MEMBRANES. Antepartum management of women with signs and symptoms of preterm labor and intact mambranes is much the same as already described for pregnancies with preterm ruptured membranes. That is, the cornerstone of treatment is to avoid delivery prior to 34 weeks' gestation if possible. Drugs intended to abate or suppress uterine contractions are commonly employed, and these will be discussed later. As in pregnancies with preterm ruptured membranes, antimicrobials, for the purpose of delaying delivery in women with preterm labor, have been studied specifically in women with intact membranes. Results with a variety of antimicrobial agents have been disappointing (Cox, 1996a; Gordon, 1995; Klebanoff, 1995, Romero, 1993, and their co-workers). Two smaller trials of metronidazole and ampicillin in women with suspected spontaneous preterm labor have shown a modest benefit (Norman and associates, 1994; Svare and colleagues, 1997). It is likely that administration of antimicrobials after preterm labor has begun is too late to interfere with propagation of the biochemical cascade that modulates uterine activity (Fig. 27-8).

GLUCOCORTICOID THERAPY. On the basis of previous observations that corticosteroids administered to the ewe accelerated lung maturation in the preterm fetus, Liggins and Howie (1972) performed a randomized study to evaluate the effects of maternally administered betamethasone (12 mg intramuscularly in two doses, 24 hours apart) to prevent respiratory distress in the subsequently delivered preterm infant. Infants born before 34 weeks had a significantly lowered incidence of respiratory distress and neonatal mortality from hyaline membrane disease if birth was delayed for at least 24 hours after completion of 24 hours of betamethasone given to the mother and for up to 7 days after completion of steroid therapy. This study served to stimulate more than 20 years of research on fetal maturation, which culminated in 1995 with a position statement on usage of corticosteroids for fetal therapy by the National Institutes of Health (NIH Consensus Development Panel, 1995).

The mechanism by which betamethasone or other corticosteroids are currently thought to reduce the frequency of respiratory distress involves induction of proteins that regulate biochemical systems within type II cells in the fetal lung that produce surfactant (Ballard and Ballard, 1995). The reported physiological effects of glucocorticoids on the developing lungs include increased alveolar surfactant, compliance, and maximal lung volume.

COLLABORATIVE STUDY ON ANTENATAL STEROID THERAPY. The NIH²-sponsored trial by the Collaborative Group on Antenatal Steroid Therapy (1981) was designed to study the effects of corticosteroid therapy on preterm fetuses. This double-blind collaborative trial was conducted at five centers, and of nearly 8000 women identified to be in preterm labor, 696 women at risk for preterm birth were enrolled. Women randomized to therapy were given intramuscular dexamethasone, 5 mg every 12 hours for a total of up to 4 doses. Including twins, a total of 720 infants was available for analysis. A significantly fewer number of infants whose mothers were given dexamethasone developed respiratory distress (13 versus 18 percent). Importantly, more than 80 percent in each study group did not develop respiratory distress. Neonatal mortality was not reduced by treatment. Similar results were obtained from a multicenter randomized trial of betamethasone performed in the United Kingdom between 1975 and 1978 (Gamsu and colleagues, 1989).

In 1985, a workshop reviewed the results of the Collaborative Study (Avery and colleagues, 1986). The group agreed that no differences in terms of cognitive, motor, or neurological function were found when 406 of the study infants were followed to 36 months of age, suggesting that steroid treatment did not adversely affect subsequent short-term neurological development (Collaborative Group, 1984). They further concluded that dexamethasone appeared beneficial only to a female fetus of older than 30 weeks and when the treatment-to-delivery interval exceeded 24 hours. Such therapy was not helpful in multiple gestations. These many caveats concerning the benefits (or lack thereof) of corticosteroids undoubtedly limited widespread use of this fetal therapy (Jobe and colleagues, 1993).

NATIONAL INSTITUTE OF HEALTH CONSENSUS DEVELOPMENT CONFERENCE ON CORTICOSTEROIDS FOR FETAL MATURATION. The proceedings of this conference were described in detail in the American Journal of Obstetrics and Gynecology (NIH Consensus Development Conference Statement, 1995). The resulting consensus on prescribing corticosteroids to accomplish fetal maturation was hailed by some as the most important development in perinatal care during 1995 (Hayward and Diaz-Rossello, 1995). Grimes (1995) took the position that because of the delay between 1972 and 1995 in adopting corticosteroids for fetal maturation, "tens of thousands of neonates died needlessly ... and our nation squandered millions of health care dollars treating preventable complications of prematurity."

The centerpiece used by the National Institutes of Health panel on corticosteroids to establish the efficacy of this therapy was a meta-analysis of randomized controlled trials (Crowley, 1995). Such analyses pool individual randomized controlled trials together to arrive at an overall estimate of the effect of an intervention, in this case, corticosteroids. This analytical approach assumes that the pooled studies are identical or nearly so in experimental design, definitions of disease end-points, patient populations, and treatment regimens. It is likely that most current meta-analyses have several sources of between-trial heterogeneity that could skew interpretation of pooled results (Moher and Olkin, 1995). For example, some trials included could use correctly concealed treatment allocation to randomize subjects whereas others may not be this rigorous in design. This is crucial, because trials with inadequate concealment generally produce results indicating that a particular treatment or intervention is beneficial (Martyn, 1996). An example of this type of heterogeneity can be found in the corticosteroid meta-analysis. Morales and colleagues (1986), in their randomized trial of corticosteroids for fetal maturation in 245 pregnancies with preterm ruptured membranes, used hospital medical record numbers in an unspecified design to randomize patients. This method of allocation to treatment or nontreatment with corticosteroids was thus not concealed from the clinicians. Despite this, the study by Morales and colleagues (1986) was pooled in the meta-analysis with other trials with more rigorous experimental design.

The next three figures include examples of meta-analysis on the fetal effects of corticosteroids using respiratory distress syndrome (Fig. 27-12), neonatal death (Fig. 27-13), and the summary of multiple effects on infant outcome (Fig. 27-14). A total of 15 randomized trials were chosen by Crowley (1995) to be included in the meta-analysis. Each horizontal line represents the results of one trial and the shorter the line, the more certain the result. If a horizontal line (study trial) touches the solid vertical line it means that that particular trial found no significant benefit. The total at the bottom of each figure is the result when all 15 trials are pooled together.

In the case of respiratory distress syndrome, nine trials found no benefit for corticosteroids and three others found only marginal benefits (Fig. 27-12). The total effect, however, when all the trials are pooled indicates that corticosteroids reduce respiratory distress by 50 percent. This beneficial effect is heavily dependent on the largest and original trial by Liggins and Howie (1972). An alternative interpretation could also be that no group of investigators has ever been able to reproduce that original trial. For neonatal deaths, 10 of the 15 trials analyzed showed no benefit (Fig. 27-13). Pooling of the data, however, indicates that corticosteroids reduced neonatal mortality by 50 percent. Meta-analysis for multiple infant outcomes related to corticosteroid therapy is shown in Figure 27-14.

Table 27-9 summarizes recommendations for use of corticosteroids for fetal maturation by the NIH Consensus Development Panel (1995). Data were deemed insufficient to assess effectiveness of corticosteroids in pregnancies complicated by hypertension, diabetes, multiple gestation, fetal growth restriction, and fetal hydrops. The Consensus Panel concluded, however, that in the absence of evidence of adverse effects, it is reasonable to use corticosteroids with these complications. It was also concluded, however, that there were insufficient data to comment on the benefits or risks of repeating maternal corticosteroid treatment weekly to enhance fetal maturation. The national response to dissemination of the NIH² Consensus Panel recommendations was dramatic, and the use of corticosteroids to enhance fetal maturation doubled within 12 months (Leviton and colleagues, 1999).

The American College of Obstetricians and Gynecologists Committee on Obstetric Practice (1994) supported the conclusions of the NIH² Consensus Panel with the exception of the recommendation for treatment of women with preterm ruptured membranes. They recommended further research. Indeed, Chapman and colleagues (1996) concluded that corticosteroid treatment for women with preterm ruptured membranes who were delivered of infants weighing less than 1000 g was not beneficial. More recently, the American College of Obstetricians and Gynecologists Committee on Obstetric Practice (1998) expressed concern about adverse fetal effects and possible effects on maternal immune status of repeated weekly courses of steroids. The College recommended that following the initial course of corticosteroids, repeated doses should be given only on an as-needed basis (i.e., a "rescue" dose given when the threat of preterm delivery recurred).

In the previous edition of this textbook, we predicted that "It is likely that some of the many benefits attributed to corticosteroid therapy by the Consensus Panel will be reassessed in the future." Indeed, the Consensus Panel was reconvened on August 17-18, 2000 to reconsider the safety and efficacy of repeated courses of corticosteroids given for fetal maturation. The conclusions of the panel are shown in Table 27-10. To summarize, it is now recommended that repeated courses of antenatal corticosteroids be used only in clinical trials.

ADVERSE EFFECTS OF CORTICOSTEROIDS. Studies initiated in the 1970s, which followed the development of children treated antenatally with corticosteroids up to the age of 12 years, showed no adverse outcomes in the areas of long-term neurodevelopment. These were measured by learning, behavioral, and motor or sensory disturbances (NIH Consensus Development Panel, 1995). There are, however, short-term maternal effects to include pulmonary edema, infection, and more difficult glucose control in diabetic women. No long-term adverse maternal effects have been reported.

Liggins and Howie (1972) based their use of corticosteroids to promote fetal lung maturation on experiments in sheep that indicated that such therapy not only affected lung maturation but also stimulated labor. Corticosteroids were reported to induce labor in humans more than 20 years ago (Jenssen and Wright, 1977; Mati and colleagues, 1973). Elliott and Radin (1995) confirmed that corticosteroids induce uterine contractions and preterm labor in humans.

Adverse fetal-newborn and long-term effects have been reported in regard to repeated versus single courses of maternal corticosteroid treatment to enhance pulmonary maturation. Esplin and colleagues (2000) compared mental and psychomotor development of 429 low-birthweight infants exposed to two courses or more of antenatal corticosteroids with infants exposed to a single course or to no exposure. The authors found no benefits for repeated doses, and exposure to multiple courses of corticosteroids was independently and significantly associated with abnormal psychomotor development. Vermillion and colleagues (2000), in an analysis of 453 infants, determined that early-onset neonatal sepsis, chorioamnionitis, and neonatal death were significantly associated with multiple maternal dosing with betamethasone. Thorp (2000) and Guinn (2001) and their associates conducted large prospective trials and found no benefits of multiple steroid courses. Importantly, in a secondary analysis, Thorp and co-workers (2001) reported a significant reduction in head circumference in infants exposed to steroids. Likewise, Mercer and associates (2001) reported a dose-dependent decrease in birth weight and length in neonates exposed to antenatal steroid therapy. Some of these results were available and they played a prominent role in the conclusions reached by the reconvened Consensus Panel (Table 27-10).

USE OF CORTICOSTEROIDS AT PARKLAND HOSPITAL. As a result of the original Consensus Panel recommendations, dexamethasone, 5 mg intramuscularly every 12 hours for four doses every 7 days, was introduced at Parkland Hospital in May 1994 for selected women at risk for preterm birth between 24 and 34 weeks. Corticosteroids were not used for fetal maturation prior to this date. Outcomes of 370 singleton pregnancies ending between 24 and 34 weeks during 12 months before dexamethasone therapy was begun were compared with outcomes of 370 similar pregnancies in which dexamethasone therapy was subsequently given. We were unable to demonstrate much benefit, if any, in this before-after population based study. Publication of this experience has not been possible.

THYROTROPIN-RELEASING HORMONE FOR FETAL MATURATION. Knight and colleagues (1994) from New Zealand reported that administration of thyrotropin-releasing hormone (400 mg intravenously) in addition to betamethasone augmented fetal lung maturation compared with betamethasone used alone. This effect is based on experimental observations that tri-iodothyronine enhances surfactant synthesis. Crowther and co-workers (1995) and the Australian Collaborative Study Group randomized 1234 women to receive thyrotropin-releasing hormone in addition to corticosteroids or to corticosteroids alone and were unable to reproduce these beneficial results. Indeed, the incidence of respiratory disease was increased in the thyrotropin-treated group! In this ACTOBAT study, the investigators also observed that 7 percent of the mothers became overtly hypertensive as a result of thyrotropin therapy. They concluded that thyrotropin-releasing hormone given to augment fetal maturation "is associated with maternal and perinatal risks and cannot be recommended." Ballard and associates (1998) enrolled 996 women before 30 weeks' gestation in a multicenter randomized double-blind, placebo-controlled trial of maternal thyrotropin-releasing hormone treatment to decrease neonatal pulmonary morbidity. All women also received betamethasone treatment. The primary outcome was chronic lung disease or infant death before 28 days. Primary and secondary outcomes were almost identical in both groups. They concluded that antenatal administration of thyrotropin-releasing hormone was not beneficial.

ANTENATAL PHENOBARBITAL AND VITAMIN K THERAPY. As reviewed by Thorp and colleagues (1995), several studies have suggested that antenatal phenobarbital and vitamin K given to the mother may reduce the incidence of intracranial hemorrhage. They randomized 272 women at risk for preterm birth to placebo or treatment with phenobarbital and vitamin K and found that such therapy did not reduce the frequency or severity of neonatal intracranial hemorrhage.

CERCLAGE. Prophylactic cervical cerclage, which is typically recommended in the United States for women with a supporting history and recurrent midtrimester losses, has also been used in Europe to prevent preterm birth. Two randomized trials of cerclage included more than 700 women at risk for preterm delivery, and neither study showed a benefit (Lazar and colleagues, 1984; Rush and associates, 1984). Prophylactic cerclage in twin pregnancies has also been shown to be of no benefit in a randomized trial (Dor and associates, 1982). The Medical Research Council of the Royal College of Obstetricians and Gynaecologists (1993) studied 1292 women from 12 countries with heterogenous and often unclear indications for cerclage to assess if this procedure prolonged pregnancy. Approximately 75 percent of women enrolled in this randomized study had previously delivered preterm infants. In 647 women, cervical sutures were placed at about 16 weeks and their outcomes compared with 645 women randomized to no cerclage. A small—from 17 to 13 percent—but significant decrease in births before 33 weeks was observed in women undergoing cerclage. Importantly, there was no difference in neonatal death between the two groups. The use of cerclage, however, was linked to increased interventions such as tocolysis and hospital admission. The investigators concluded that cervical sutures should be offered to women with a history of three or more pregnancies ending before 37 weeks.

Althuisius and co-workers (2000) randomized 73 women who had had a prior preterm birth before 34 weeks' gestation either to cerclage or observation. These latter women were followed with transvaginal measurement of cervical length until 27 weeks, and cerclage was placed if the cervical length shortened to less than 25 mm. Delivery before 34 weeks and neonatal survival were similar in the two study groups. In a subgroup analysis, however, preterm birth was significantly reduced in women who received cerclage after developing ultrasound evidence of cervical shortening. Rust and colleagues (2000) randomized 61 women to cerclage or to expectant management based on ultrasound evidence of cervical funneling between 16 to 24 weeks' gestation. No benefits were apparent in the women randomized to cervical cerclage.

BACTERIAL VAGINOSIS. Hillier and colleagues (1995) detected bacterial vaginosis in 16 percent of 10,397 pregnant women without risk factors for preterm birth who were screened between 23 and 26 weeks. Bacterial vaginosis was linked to poverty and to significantly increased preterm birth at 32 weeks or less but not to premature membrane rupture. Microbes most strongly associated with bacterial vaginosis included Gardnerella vaginalis, Bacteroides species, and Mycoplasma hominis.

Hauth and co-workers (1995) studied 624 women at risk for preterm birth either because they had previously delivered preterm infants or because they weighed less than 50 kg before pregnancy. In this double-blind study, women were randomized to treatment with metronidazole, 250 mg three times a day for 7 days; erythromycin, 333 mg three times a day for 14 days; or placebo. About 40 percent of women in each arm had bacterial vaginosis at 22 to 24 weeks. Delivery before 37 weeks in women with bacterial vaginosis was significantly decreased from 40 to 25 percent when metronidazole or erythromycin was given. Morales and co-workers (1994) reported a similar benefit in 80 women at increased risk for preterm birth and with bacterial vaginosis and who were treated with metronidazole.

Carey and colleagues (2000), in a multicenter NICHD¹ Maternal-Fetal Medicine Units Network trial, found that metronidazole treatment of asymptomatic pregnant women with bacterial vaginosis did not improve pregnancy outcomes despite that bacterial vaginosis resolved in 78 percent of women assigned to metronidazole treatment compared with 37 percent assigned to a placebo. Specifically, the incidence of preterm birth was similar in the 953 women treated with metronidazole compared with the 966 in the placebo group. Finally, treatment with metronidazole did not reduce clinical intra-amnionic or postpartum infections, neonatal sepsis, or admission to the neonatal intensive care unit.

Intravaginal topical treatment of bacterial vaginosis using clindamycin cream has not been shown to be effective to prevent preterm birth (Joesoef and colleagues, 1995; McGregor and associates, 1994).

METHODS USED TO INHIBIT PRETERM LABOR

A great number of drugs and other interventions have been used to inhibit preterm labor, but unfortunately, none has been completely effective (American College of Obstetricians and Gynecologists, 1995). Potential maternal complications of tocolytic drugs are shown in Table 27-11. Their importance cannot be underestimated. For example, tocolysis was the third most common cause of acute respiratory distress syndrome and death in pregnant women during a 14-year period in Jackson, Mississippi (Perry and associates, 1996). The American College of Obstetricians and Gynecologists (1998) has recommended that tocolysis be used in the presence of regular uterine contractions plus documented cervical change or appreciable cervical dilatation and effacement.

BED REST. The treatment regimen that has been used most often is bed rest either in the hospital or at home. Goldenberg and colleagues (1994) have reviewed bed rest used to treat a variety of pregnancy complications and found no conclusive evidence that bed rest was helpful in preventing preterm birth. Kovacevich and associates (2000) reported that enforced bed rest (except for bathroom privileges) for 3 days or more increased the risk of thromboembolic complications from 1 per 1000 women without bed rest to 16 per 1000 in those with bed rest for threatened preterm delivery.

HYDRATION AND SEDATION. Helfgott and associates (1994) performed the first randomized trial of hydration and sedation compared with bed rest alone in the treatment of 119 women in preterm labor. Women randomized to treatment received 500 mL of lactated Ringer solution intravenously over 30 minutes and 8 to 12 mg of intramuscular morphine sulfate. Such therapy was not found to be more beneficial than bed rest.

BETA-ADRENERGIC RECEPTOR AGONISTS. Earlier in this century, epinephrine in low doses was demonstrated to exert a depressant effect on pregnant myometrium. Its tocolytic effects, however, proved to be rather weak, quite transient, and likely to be accompanied by troublesome cardiovascular effects. Several compounds capable of reacting predominantly with b-adrenergic receptors have subsequently been investigated. Some of these now are used in obstetrics, but only ritodrine hydrochloride has been approved (1980) by the Food and Drug Administration to treat preterm labor.

The adrenergic receptors are located on the outer surface of the smooth muscle cell membrane, where specific agonists can couple with them. Adenyl cyclase in the cell membrane is activated by the receptor stimulation. Adenyl cyclase enhances the conversion of ATP to cyclic AMP, which in turn initiates a number of reactions that reduce the intracellular concentration of ionized calcium and thereby prevent activation of contractile proteins (also see Chap. 11, p. 271). Flier and Underhill (1996) have comprehensively reviewed adrenergic receptors.

There are two classes of b-adrenergic receptors: b₁-receptors, dominant in the heart and intestines; and b₂-receptors, dominant in the myometrium, blood vessels, and bronchioles. A number of compounds generally similar in structure to epinephrine have been evaluated in the search for one that ideally would provide optimal stimulation of myometrial b₂-receptors and thus inhibit uterine contractions while simultaneously causing few adverse effects from stimulation of receptors elsewhere. Thus far, no compound has exhibited these utopian properties. In the United States, b-agonist compounds employed to arrest preterm labor include ritodrine and terbutaline.

RITODRINE. In a multicenter United States study, infants whose mothers were treated with ritodrine for presumed preterm labor had a lower mortality rate, developed respiratory distress less often, and achieved a gestational age of 36 weeks or a birthweight of 2500 g more often than did infants whose mothers were not so treated (Merkatz and colleagues, 1980). Hesseldahl (1979), however, in a multicenter controlled study in Denmark, did not find any of several ritodrine regimens tested to be more efficacious than treatment with bed rest, glucose infusion, and placebo tablets.

Because of contemporaneous concerns for the efficacy and safety of ritodrine, we evaluated the drug on the obstetrical service at Parkland Hospital (Leveno and associates, 1986b). Preterm labor was carefully defined to include cervical dilatation plus regular uterine contractions, and 106 women between 24 and 33 weeks were randomly allocated to receive either intravenous ritodrine or no tocolysis. Although ritodrine treatment significantly delayed delivery for 24 hours or less, it did not significantly modify ultimate perinatal outcomes. A significant delay in delivery at 48 hours was found in a randomized study involving 708 pregnancies reported by the Canadian Preterm Labor Investigators Group (1992). A likely explanation for the transient uterine tocolytic effects of ritodrine and ultimate failure of such therapy may be the phenomenon of b-adrenergic receptor desensitization (Hausdorff and colleagues, 1990).

The infusion of ritodrine, as well as the other b-adrenergic agonists, has resulted in frequent and at times serious side effects (Table 27-11). Maternal tachycardia, hypotension, apprehension, chest tightness or pain, electrocardiographic S-T segment depression, pulmonary edema, and death have all been observed. Maternal metabolic effects include hyperglycemia, hyperinsulinemia (unless diabetic), hypokalemia, and lactic and ketoacidosis. Less serious but nonetheless troublesome side effects include emesis, headaches, tremulousness, fever, and hallucinations.

A single mechanism has not been identified to explain the development of pulmonary edema, but maternal infection appears to increase the risk (Hatjis and Swain, 1988). The cause of the pulmonary edema appears to be multifactorial. Beta-adrenergic agonists cause retention of sodium and water, and thus with time—usually 24 to 48 hours—may lead to volume overload (Hankins and colleagues, 1988). The drugs have also been implicated as a cause of increased capillary permeability, disturbances of cardiac rhythm, and myocardial ischemia. Simultaneous administration of glucocorticoids to try to hasten fetal maturation may also contribute, although pulmonary edema has developed in their absence.

Only parenteral ritodrine is now available in the United States since the manufacturer discontinued distribution of tablets in 1995. The efficacy of oral ritodrine had been challenged on pharmacokinetic grounds (Schiff and colleagues, 1993).

TERBUTALINE. This b-agonist is commonly used to forestall preterm labor although, like ritodrine, toxicity—especially maternal pulmonary edema and glucose intolerance—have been evident with its use (Angel and associates, 1988).

Lam and colleagues (1988) described long-term subcutaneous administration of low-dose terbutaline using a portable pump in nine pregnancies. It was claimed that the lower dose of terbutaline used likely prevented b-adrenergic desensitization, resulting in less "breakthrough tocolysis." The Tokos Corporation promptly marketed this approach, and between 1987 and 1993 had used these pumps in nearly 25,000 women with preterm labor (Perry and colleagues, 1995). The list price for terbutaline pump therapy in Dallas in 1996 was $484 per day. The only other reports concerning terbutaline pumps include a sudden maternal death (Hudgens and Conradi, 1993) and a description of newborn myocardial necrosis after the mother used the pump for 12 weeks (Fletcher and colleagues, 1991).

Two prospective randomized trials have not found any benefit for terbutaline pump therapy. Wenstrom and colleagues (1997) randomized 42 women to therapy with a terbutaline pump, a saline pump, or to oral terbutaline. These three groups had similar gestational ages at entry and at delivery. Guinn and associates (1998) in a double-blind trial, randomized 52 women to terbutaline pump or to saline pump therapy. A sample size of 48 women was required to detect a 2-week intergroup difference in mean time to delivery. Terbutaline pump therapy did not significantly prolong pregnancy, prevent preterm delivery, or improve neonatal outcomes.

Oral terbutaline therapy has been reported ineffective by several groups (How and associates, 1995; Parilla and co-workers, 1993). In a double-blind trial, Lewis and colleagues (1996) randomized 203 women with preterm labor following successful intravenous tocolysis at 24 to 34 weeks' gestation, to 5 mg oral terbutaline every 4 hours or placebo. Delivery at one week was similar in both groups as was median latency, mean gestational age at delivery, and the incidence of recurrent preterm labor. Post hoc analysis, however, of 96 women enrolled before 32 weeks' gestation suggested significant pregnancy prolongation with maintenance oral terbutaline.

OVERVIEW OF BETA-ADRENERGIC DRUGS TO INHIBIT PRETERM LABOR. Several meta-analyses of parenteral b-agonists given to prevent preterm birth have consistently confirmed that these agents delay delivery for no more than 48 hours (Canadian Preterm Labor Group, 1992). Moreover, this delay has not proven to be beneficial despite repeated attempts to revisit the data (Lamont, 1993). Finally, Macones and colleagues (1995) used meta-analysis to assess the available data on the efficacy of oral b-agonist therapy and found no benefits.

Thus, oral b-agonist therapy has convincingly been shown to be ineffective, and parenteral therapy can only delay delivery for a short time that has not been shown to be beneficial. Keirse (1995b) suggests that the brief delay in delivery afforded may be useful to facilitate maternal transport to tertiary care centers, and also delay delivery sufficiently to effect fetal maturation with glucocorticoids. Unfortunately, there are no data to support this viewpoint.

MAGNESIUM SULFATE. It has been recognized for some time that ionic magnesium in a sufficiently high concentration can alter myometrial contractility in vivo as well as in vitro. Its role is presumably that of a calcium antagonist.

Steer and Petrie (1977) concluded that intravenously administered magnesium sulfate, 4 g given as a loading dose followed by a continuous infusion of 2 g/hr, will usually arrest labor. Elliott (1983), in a retrospective study, found tocolysis with magnesium sulfate to be successful, inexpensive, and relatively nontoxic. He reported 87 percent success when the cervix was dilated 2 cm or less, but the period of inhibited labor was as short as 48 hours. Spisso and co-workers (1982) were also favorably impressed by the efficacy of magnesium sulfate when given intravenously in relatively large doses to women in the early latent phase of labor.

Watt-Morse and associates (1995) studied the inhibitory effects of magnesium concentrations up to 8.3 mEq/L in preterm sheep with oxytocin-induced contractions. They concluded that magnesium sulfate in tolerable, nontoxic doses has no direct effect on uterine contractility.

There have been only two randomized controlled studies of the tocolytic properties of magnesium sulfate in humans. Cotton and associates (1984) compared magnesium sulfate with ritodrine as well as with a placebo, and they identified little difference in outcomes. Cox and associates (1990) randomized 156 women in preterm labor with intact membranes to infusions of magnesium sulfate or normal saline. Magnesium sulfate (20 percent solution) was begun using a 4-g loading dose followed by 2 g/hr intravenously. If contractions persisted after 1 hour, the infusion was increased to 3 g/hr. Treated women had a mean plasma magnesium concentration of 5.5 mEq/L. No benefits for such therapy were found, and this method of tocolysis was abandoned at Parkland Hospital. Similar results were subsequently reported in an evaluation of nonrandomized women in preterm labor who were delivered of infants weighing less than 1000 g (Kimberlin and associates, 1996a).

Hollander and colleagues (1987) used an unprecedented infusion dose of magnesium sulfate that averaged 4.5 g/hr. They reported that such therapy was equivalent to ritodrine. Conversely, Semchyshyn and associates (1983) failed to stop labor in a woman who inadvertently was given 17.3 g of magnesium sulfate in 45 minutes! Women given high-dosage magnesium sulfate must be monitored very closely for evidence of hypermagnesemia that might prove toxic to them and their fetus-infants. The pharmacology and toxicology of parenterally administered magnesium are considered in more detail in Chapter 24 (p. 599).

PROSTAGLANDIN INHIBITORS. Compounds that inhibit prostaglandins have been the subject of considerable interest since it was appreciated that prostaglandins are intimately involved in myometrial contractions of normal labor (Chap. 11, p. 274). Antiprostaglandin agents may act by inhibiting the synthesis of prostaglandins or by blocking the action of prostaglandins on target organs. A group of enzymes collectively called prostaglandin synthase is responsible for the conversion of free arachidonic acid to prostaglandins. Several drugs are known to block this system, including aspirin and other salicylates, indomethacin, naproxen, and sulindac.

Unfortunately, prostaglandin synthase inhibitors may adversely affect the fetus, and this has prevented widespread use of these agents for tocolysis. Complications include closure of the ductus arteriosus, necrotizing enterocolitis, and intracranial hemorrhage (Norton and co-workers, 1993). Parilla and colleagues (2000) have challenged the association of indomethacin with necrotizing enterocolitis. Van der Heijden and colleagues (1994) linked long-term perinatal indomethacin to anemia, neonatal death, and cystic renal damage. The mother can be adversely affected by indomethacin therapy, and Lunt and associates (1994) reported that indomethacin tocolysis caused profound prolongation of maternal bleeding time.

Sulindac, closely related to indomethacin in structure, has been reported to have fewer side effects when used for tocolysis (Rasanen and Jouppila, 1995). Preliminary trials, however, indicate that oral sulindac therapy may not be very useful in the prevention of preterm birth (Carlan and associates, 1995). Kramer and colleagues (1996) measured the effects of sulindac on fetal urine production and amnionic fluid volume and compared them with terbutaline in a randomized, double-blind study. Sulindac administration decreased fetal urine flow and amnionic fluid volume. Two fetuses also developed severe ductal constriction. Thus, sulindac shares many of the fetal side effects associated with indomethacin.

Panter and colleagues (1996) reviewed all randomized trials that have compared indomethacin with b-agonists for tocolysis. Indomethacin was found to be more effective in delaying delivery by 48 hours, and there were fewer maternal side effects compared with ritodrine. Indomethacin was, however, associated with increased neonatal morbidity. These investigators concluded that indomethacin needs to be further evaluated before it is used routinely for tocolysis.

CALCIUM-CHANNEL-BLOCKING DRUGS. Smooth muscle activity, including myometrium, is directly related to free calcium within the cytoplasm, and a reduction in calcium concentration inhibits contraction. Calcium ions reach the cytoplasm through specific membrane portals or channels, and calcium-channel blockers act to inhibit, by a variety of different mechanisms, the entry of calcium through the cell membrane channels. Calcium-entry blockers, because of their smooth muscle arteriolar relaxation effects, are currently being used for the treatment of coronary artery disease and hypertension.

The possibility that calcium-channel-blocking drugs might have applications in the treatment of preterm labor has been the subject of research in both animals and humans since the late 1970s. Saade and colleagues (1994), using in vitro human myometrial strips, showed that nifedipine caused relaxation similar to ritodrine and more effectively than magnesium. The first clinical trial in which nifedipine was given for preterm labor was from Denmark by Ulmsten and colleagues (1980). Nifedipine treatment postponed delivery at least 3 days in 10 women with preterm labor at 33 weeks or less. No serious maternal or fetal side effects were noted. There have been several subsequent studies of nifedipine tocolysis, and these have been reviewed comprehensively by Childress and Katz (1994). In all the studies, nifedipine was as successful or better than ritodrine in stopping preterm contractions without adverse fetal effects. Maternal side effects were much worse with ritodrine.

As promising as calcium-channel blockers may appear for treatment of preterm labor, some investigators caution that more research is needed to clarify their potential maternal or fetal dangers. This is because smooth muscle relaxation by nifedipine is not limited to uterine muscle, but also includes the systemic and uterine vasculature. Nifedipine-induced decreased vascular resistance can lead to maternal hypotension and thus decreased uteroplacental perfusion. Parisi and colleagues (1986) reported that hypercapnia, acidosis, and possibly hypoxemia developed in fetuses of hypertensive ewes given nicardipine. Similarly, Lirette and colleagues (1987) observed a fall in uteroplacental blood flow in pregnant rabbits. Other investigators, however, have not found these adverse fetal effects (Childress and Katz, 1994).

The efficacy of calcium-channel blockers, such as nifedipine, in suppressing preterm contractions or labor has not been adequately studied. Small trials have arrived at inconclusive results (Kupferminc and colleagues, 1993; Papatsonis and co-workers, 1997; Read and associates, 1986; Smith and Woodland, 1993). In a review from the Cochrane Database comparing nifedipine and b-sympathomimetics, Keirse (1995a) concluded that nifedipine treatment reduced births less than 2500 g, but more infants were admitted to neonatal intensive care. Results such as these mandate further study of their safety and efficacy.

The combination of nifedipine and magnesium for tocolysis is potentially dangerous. Ben-Ami and colleagues (1994) and Kurtzman and associates (1993) have reported that nifedipine enhances the toxicity of magnesium to produce neuromuscular blockade that can interfere with both pulmonary and cardiac function. Carr and colleagues (1999) found that maintenance therapy with oral nifedipine did not significantly prolong pregnancy in women initially treated with intravenous magnesium sulfate for preterm labor.

ATOSIBAN. A nonapeptide oxytocin analog, atosiban has been shown to be a competitive oxytocin-vasopressin antagonist capable of inhibiting oxytocin-induced uterine contractions. Goodwin and colleagues (1995) reviewed atosiban and described its pharmacokinetics in pregnant women. Results of a multicenter, double-blind, placebo-controlled trial in 501 women assigned to intravenous atosiban or placebo were recently published (Romero and colleagues, 2000). Atosiban therapy did not significantly improve any clinically relevant infant outcome. Safety considerations were of great concern with more fetal-infant deaths in the atosiban group. Food and Drug Administration approval for the use of atosiban to arrest preterm labor has been denied due to concerns regarding efficacy and fetal-newborn safety (FDA, personal communication).

NITRIC OXIDE DONOR DRUGS. Nitric oxide is a potent endogenous smooth-muscle relaxant in the vasculature, the gut, and the uterus. Nitroglycerin is an example of a nitric oxide donor drug. Clavin and colleagues (1996) randomized 34 women in preterm labor to tocolysis with intravenous nitroglycerin or magnesium sulfate. There was no difference in the tocolytic efficacy of these two drugs, but 3 of 15 women given nitroglycerin had severe hypotension. Lees and colleagues (1999) randomized 245 women in preterm labor to transdermal glycerl trinitrate or intravenous ritodrine. There was no overall superiority of glyceryl trinitrate in the inhibition of preterm labor.

COMBINED THERAPY. The use of multiple drugs to inhibit preterm labor suggests that no single drug is completely satisfactory. Unfortunately, there are no studies in which tocolytic combinations were compared with placebo. Kosasa and colleagues (1994) used long-term tocolysis with combined intravenous terbutaline and magnesium sulfate in 1000 pregnancies. The mean duration of intravenous therapy was 61 days in women with intact membranes, and one woman received this therapy for 123 days. This approach was claimed to be effective and safe because only 2 to 4 percent of the women developed pulmonary edema.

INTRAPARTUM MANAGEMENT

In general, the more immature the fetus, the greater the risks from labor and delivery.

LABOR. Whether labor is induced or spontaneous, abnormalities of fetal heart rate and uterine contractions should be sought, preferably by continuous electronic monitoring. Fetal tachycardia, especially in the presence of ruptured membranes, is suggestive of sepsis. There is some emerging evidence that intrapartum acidemia may intensify some of the neonatal complications usually attributed solely to prematurity. For example, Low and colleagues (1995) observed that intrapartum acidosis—umbilical artery blood pH less than 7.0—had an important role in neonatal complications. Similarly, Kimberlin and colleagues (1996b) found that increasing umbilical artery blood acidemia was related to more severe respiratory disease in preterm neonates, although no effects were found in short-term neurological outcomes that included intracranial hemorrhages.

Importantly, just as the markedly preterm infant is to be afforded special care in the neonatal intensive care unit, the mother and fetus should be observed very closely in the labor and delivery unit. Especially skilled physicians should monitor the labor and delivery of the markedly preterm fetus.

PREVENTION OF NEONATAL GROUP B STREPTOCOCCAL INFECTIONS. As discussed in Chapter 56 (p. 1471), group B streptococcal infections are common and dangerous in the preterm neonate. Since 1996 the Centers for Disease Control and Prevention, along with the American College of Obstetricians and Gynecologists, recommend either penicillin G or ampicillin intravenously every 6 hours until delivery for women in labor prior to 37 weeks and whose culture status is unknown or positive for group B streptococcus.

DELIVERY. In the absence of a relaxed vaginal outlet, an episiotomy for delivery may be advantageous once the fetal head reaches the perineum. Argument persists as to the merits of spontaneous delivery versus forceps delivery to protect the fragile preterm fetal head (Chap. 21, p. 490). It is doubtful whether use of forceps in most instances produces less trauma. Indeed, to compress and pull on the head of a grossly preterm infant might be more traumatic than natural expulsion. The use of outlet forceps of appropriate size may be of assistance when conduction analgesia is used and voluntary expulsion efforts are obtunded.

A physician and staff proficient in resuscitative techniques and fully oriented to the specific problems of the case should be present at delivery. Principles of resuscitation described in Chapter 16 are applicable, including prompt tracheal intubation and ventilation. The importance of the availability of specialized personnel and facilities in the case of preterm infants is underscored by the improved survival of these infants when they are delivered in tertiary care centers (Powell and colleagues, 1995).

PREVENTION OF NEONATAL INTRACRANIAL HEMORRHAGE. Preterm infants frequently have germinal matrix bleeding that can extend to more serious intraventricular hemorrhage (Bejar and colleagues, 1980). It was hypothesized that cesarean delivery to obviate trauma from labor and vaginal delivery might prevent these complications. These initial observations have not been validated by most subsequent studies. In the largest study, Malloy and colleagues (1991) analyzed 1765 infants with birthweights less than 1500 g and found that cesarean delivery did not lower the risk of mortality or intracranial hemorrhage. Anderson and colleagues (1988), however, made an interesting observation regarding the role of cesarean delivery in the prevention of neonatal intracranial hemorrhages. These hemorrhages were related to whether or not the fetus had been subjected to the active phase of labor, defined as the interval before 5 cm cervical dilatation. As emphasized by Anderson and colleagues (1988), avoidance of active-phase labor is impossible in most preterm births because the route of delivery cannot be decided until labor is firmly established.

Nelson and Grether (1995) reported that magnesium sulfate given to women delivered preterm for either tocolysis or preeclampsia was associated with a significantly reduced incidence of cerebral palsy when surviving infants with birthweights less than 1500 g were followed to 3 years of age. It was suggested that magnesium given to the fetus via the mother perhaps played a role in regulation of the vasculature supplying the germinal matrix of the preterm fetal brain that is especially vulnerable to hemorrhage. Murphy and colleagues (1995), however, found that severe preeclampsia and delivery without labor were protective against cerebral palsy. They concluded that magnesium could not be the protective agent because this drug is not used for preeclampsia in England.