From Williams Obstetrics:

25. Obstetrical Hemorrhage

Obstetrics is "bloody business." Even though the maternal mortality rate has been reduced dramatically by hospitalization for delivery and the availability of blood for transfusion, death from hemorrhage remains prominent in the majority of mortality reports in advanced countries. In the United States from 1979 through 1992, the Centers for Disease Control and Prevention analyzed 4915 nonabortion-related maternal deaths from the Pregnancy Mortality Surveillance System (Chichakli and colleagues, 1999). They found that hemorrhage was a direct cause in about 30 percent of these deaths. According to Bonnar (2000), hemorrhage was the major factor in maternal deaths in the United Kingdom between 1985 and 1996. There undoubtedly has been great improvement in mortality from hemorrhage with modernization of American obstetrics. For example, Sachs and associates (1987) reported that maternal deaths from obstetrical hemorrhage in Massachusetts decreased tenfold from the mid-1950s to the mid-1980s.

Unfortunately, despite improved outcomes, poor and minority women continue to die from hemorrhage and its complications at a disparately high rate. In the report from the Centers for Disease Control cited above, there was threefold increased mortality from hemorrhage in African-American compared with Caucasian women. In a similar analysis of 3777 pregnancy-related deaths from states that include Hispanic origin on death certificates, Hopkins and co-workers (1999) reported that hemorrhage caused 20 percent of maternal deaths. They showed disparate mortality in African-American and Hispanic women compared with Caucasians.

Causes of maternal death from hemorrhage are shown in Table 25-1. Obstetrical hemorrhage is most likely to be fatal in circumstances in which blood or components are not available immediately. The establishment and maintenance of facilities that allow prompt administration of blood are absolute requirements for acceptable obstetrical care. Hemorrhage may be antepartum—such as with placenta previa or placental abruption, or it more likely develops postpartum—from uterine atony or genital tract lacerations.

INCIDENCE AND PREDISPOSING CONDITIONS. Because of inexact definitions used, the incidence of obstetrical hemorrhage cannot be determined precisely. In one study of women delivered vaginally, Combs and colleagues (1991b) defined hemorrhage by a postpartum hematocrit drop of 10 volume percent or by need for transfusion. Using these criteria, the incidence was 3.9 percent. In women undergoing cesarean delivery it was 6 to 8 percent (Combs and associates, 1991a; Naef and co-workers, 1994).

Dickason and Dinsmoor (1992) reported transfusions in 6.8 percent of women undergoing cesarean delivery. Klapholz (1990) reviewed over 30,000 deliveries at the Beth Israel Hospital from 1976 to 1986. As expected, the incidence of transfusion has decreased over the years; in 1976 it was 4.6 percent, but by 1986 it was 1.9 percent.

Table 25-2 lists the many clinical circumstances in which risk of hemorrhage is appreciably increased. It is apparent that serious hemorrhage may occur at any time throughout pregnancy and the puerperium. The time of bleeding in pregnancy is widely used to classify obstetrical hemorrhage; however, the term third-trimester bleeding is imprecise and its use not recommended. One factor not generally considered as "predisposing" to hemorrhagic death is the lack of availability of obstetrical and anesthetic services. According to Bonnar (2000), the majority of deaths from hemorrhage in the United Kingdom cited earlier were associated with substandard care. Similarly, Nagaya and associates (2000) reviewed 197 maternal deaths in Japan in the 2-year period spanning 1991 and 1992. Hemorrhage caused 40 percent of these deaths, and they concluded that many of these were preventable because they were associated with inadequate obstetrical facilities. In Japan, 40 percent of deliveries take place in clinics with less than 20 beds, and frequently the physician functions both as obstetrician and anesthesiologist.

ANTEPARTUM HEMORRHAGE

Slight vaginal bleeding is common during active labor. This "bloody show" is the consequence of effacement and dilatation of the cervix, with tearing of small veins. Uterine bleeding from a site above the cervix before delivery is cause for concern. The bleeding may be the consequence of some separation of a placenta implanted in the immediate vicinity of the cervical canal—placenta previa. It may come from separation of a placenta located elsewhere in the uterine cavity—abruptio placentae. Rarely, the bleeding may be the consequence of velamentous insertion of the umbilical cord with rupture and hemorrhage from a fetal blood vessel at the time of rupture of the membranes—vasa previa.

The source of uterine bleeding that originates above the level of the cervix is not always identified. In that circumstance, the bleeding typically begins with little or no other symptomatology, and then stops, and at delivery no anatomical cause is identified. Almost always the bleeding must have been the consequence of slight marginal separation of the placenta that did not expand. The pregnancy in which such bleeding occurs remains at increased risk for a poor outcome even though the bleeding soon stops and placenta previa appears to have been excluded by sonography. Lipitz and colleagues (1991) studied 65 consecutive women—almost 1 percent of their patients—who had uterine bleeding between 14 and 26 weeks. Almost a fourth had placental abruption or previa. Total fetal loss including abortions and perinatal deaths was 32 percent. Even in pregnancies with hemorrhage after 26 weeks that are not explained by placental abruption or previa, Ajayi and colleagues (1992) reported adverse outcomes in a third. For this reason, delivery should be considered in any woman at term with unexplained vaginal bleeding.

PLACENTAL ABRUPTION

The separation of the placenta from its site of implantation before the delivery of the fetus has been variously called placental abruption, abruptio placentae, and in Great Britain, accidental hemorrhage. The term premature separation of the normally implanted placenta is most descriptive because it differentiates the placenta that separates prematurely but is implanted some distance beyond the cervical internal os, from one that is implanted over the cervical internal os—that is, placenta previa. It is cumbersome, however, and hence the shorter term abruptio placentae, or placental abruption, has been employed. The Latin abruptio placentae, which means "rending asunder of the placenta," denotes a sudden accident, a clinical characteristic of most cases of this complication.

Some of the bleeding of placental abruption usually insinuates itself between the membranes and uterus, and then escapes through the cervix, causing external hemorrhage (Fig. 25-1). Less often, the blood does not escape externally but is retained between the detached placenta and the uterus, leading to concealed hemorrhage (Figs. 25-1 and 25-2). Placental abruption may be total (Figs. 25-1 and 25-2) or partial (Fig. 25-3). Placental abruption with concealed hemorrhage carries with it much greater maternal hazards, not only because of the possibility of consumptive coagulopathy, but also because the extent of the hemorrhage is not appreciated.

FREQUENCY, INTENSITY, AND SIGNIFICANCE. The frequency with which abruptio placentae is diagnosed will vary because criteria employed for diagnosis differ. The intensity of the abruption will often vary depending on how quickly the woman seeks and receives care following the onset of symptoms. With delay, the likelihood of extensive separation causing death of the fetus is increased remarkably.

The reported frequency for placental abruption averages about 1 in 200 deliveries. Karegard and Gennser (1986) surveyed 849,619 births in Sweden and reported that 1 in 225 were complicated by abruptio placentae. Ananth and colleagues (1999a) reviewed 13 studies with nearly 1.6 million pregnancies and reported an incidence of 1 in 155.

At Parkland Hospital from 1988 through 1999, the incidence of abruption in over 169,000 deliveries has been 1 in 290. The incidence as well as severity have decreased over time. Applying the criterion of placental separation so extensive as to kill the fetus, the incidence was 1 in 420 deliveries from 1956 through 1967 (Pritchard and Brekken, 1967). As the number of high-parity women cared for decreased, and community-wide availability of prenatal care as well as emergency transportation improved, the frequency of abruption causing fetal death has dropped to about 1 in 830 deliveries from 1974 through 1989 (Pritchard and colleagues, 1991). From 1988 through 1999, it decreased to 1 in 1550.

PERINATAL MORBIDITY AND MORTALITY. In most reports, perinatal mortality with placental abruption is about 25 percent. In the large Swedish study by Karegard and Gennser (1986) cited earlier, it was 20 percent. Krohn and associates (1987) reported that perinatal mortality was 20 percent in 884 pregnancies complicated by placental abruption in Washington State. Ananth and co-workers (1999b) studied 530 women with placental abruption at Mt. Sinai Hospital in New York and reported that 40 percent were delivered preterm.

As stillbirths from other causes have decreased appreciably, those from abruptio placentae have become especially prominent. For example, of all third-trimester stillbirths in over 40,000 deliveries at Parkland Hospital during 1992 and 1994, 12 percent were the consequence of placental abruption (Cunningham and Hollier, 1997). This frequency is similar to that described by Fretts and Usher (1997), who studied almost 62,000 births at the Royal Victoria Hospital in Montreal between 1978 and 1995. Abruptio placentae had become the leading known cause and accounted for 15 percent of stillborns.

Importantly, even if the infant survives, there may be adverse sequelae. Of the 182 survivors in the study by Abdella and associates (1984), 25 (14 percent) were identified to have significant neurological deficits within the first year of life.

ETIOLOGY. The primary cause of placental abruption is unknown, but there are several associated conditions. Some of these are listed in Table 25-3. As shown in Figure 25-4, the incidence increases with maternal age. While Pritchard and colleagues (1991) have also shown it to be higher in women of great parity, Toohey and associates (1995) did not find this in women para 5 or greater. Race or ethnicity appears to be important. In the 169,000-plus deliveries at Parkland Hospital, abruption was more common in African-American and Caucasian women (1 in 200) than in Asian (1 in 300) or Latin-American women (1 in 450).

By far the most commonly associated condition is some type of hypertension. This includes preeclampsia, gestational hypertension, or chronic hypertension. In the earlier Parkland Hospital study of 408 cases of placental abruption so severe as to kill the fetus, maternal hypertension was apparent in about half of the women once the depleted intravascular compartment was adequately refilled (Pritchard and co-workers, 1991). Half of these women had chronic hypertension and the remainder had either gestational hypertension or preeclampsia. Morgan and colleagues (1994) found that hypertensive women were more likely to suffer a more severe abruption. According to Witlin and colleagues (1999), however, the severity of preeclampsia did not correlate with the incidence of abruption in 445 women. From the Maternal-Fetal Medicine Network, Sibai and co-workers (1998) reported that 1.5 percent of women with chronic hypertension suffered placental abruption. Ananth and associates (1999b) reported a threefold increased incidence of abruption with chronic hypertension and fourfold with severe preeclampsia.

There is an increased incidence of abruption with preterm prematurely ruptured membranes. Gonen and associates (1989) reported an incidence of 5.6 percent in 143 pregnancies of less than 34 weeks in which the membranes were ruptured for more than 24 hours. Major and colleagues (1995) described an incidence of 5 percent in 756 women with ruptured membranes between 20 and 36 weeks. Kramer and co-workers (1997) found an incidence of 3.1 percent in all patients if membranes were ruptured for longer than 24 hours. In a meta-analysis of 54 studies, Ananth and colleagues (1996) found a threefold risk of abruption with prematurely ruptured membranes.

In the earliest studies, from the Collaborative Perinatal Project, cigarette smoking was linked to an increased risk for abruption (Misra and Ananth, 1999; Naeye, 1980). In their meta-analysis of 1.6 million pregnancies, Ananth and colleagues (1999a, 1999b) found a twofold risk for abruption in smokers. This was increased to five- to eightfold if smokers had chronic hypertension and/or severe preeclampsia.

Cocaine abuse has been associated with an alarming frequency of placental abruption. In one report of 50 women who abused cocaine during pregnancy, there were eight stillbirths caused by placental abruption (Bingol and associates, 1987). Hoskins and collaborators (1991) reported a 13-percent rate of placental abruption in 112 women who were followed prospectively for cocaine abuse during pregnancy. Slutsker (1992) reviewed 10 studies of cocaine-using women and all showed that placental abruption was more common than in controls.

Over the past decade, a number of inherited or acquired thrombophilias have been described that are associated with thromboembolic disorders during pregnancy (Chap. 46, p. 1234). These clotting disorders also are associated with placental abruption and infarction (Gherman and Goodwin, 2000). Most of these are single-gene mutations that include genes for factor V Leiden, prothrombin, methylenetetrahydrofolate reductase, proteins S and C, and antithrombin III (Chap. 49, p. 1330). In addition, acquired antiphospholipid autoantibodies, including the lupus anticoagulant, are associated with placental abruption. For example, Kupferminc and colleagues (1999) found a sevenfold risk of abruption in women with a factor V, folate reductase, or prothrombin mutation.

As discussed in Chapter 43 (p. 1173), external trauma was implicated in only 3 of 207 cases of placental abruption causing fetal death at Parkland Hospital. Our experiences are similar to those of Kettel (1988) and Stafford (1988) and their co-workers, who stressed that abruption caused by relatively minor trauma may cause fetal jeopardy that is not always associated with immediate evidence for placental separation. In such cases, a period of monitoring for at least 4 hours is often necessary to exclude a subclinical abruption.

Uterine leiomyoma, especially if located behind the placental implantation site, predisposes to abruption (Chap. 35, p. 926). Rice and associates (1989) reported that 8 of 14 women with retroplacental myomas developed placental abruption; in four, fetal death ensued. By contrast, abruption developed in only 2 of 79 women whose myoma was not retroplacental.

RECURRENT ABRUPTION. Pritchard and co-workers (1970) identified a recurrence rate of severe abruption in 1 in 8 pregnancies. Importantly, of the 14 recurrent placental abruptions, eight caused fetal death for a second time. Karegard and Gennser (1986) reported that the recurrent placental abruption risk was increased tenfold, from 0.4 to 4 percent. Management of the subsequent pregnancy is made difficult in that the placental separation may suddenly occur at any time, even remote from term. In the majority of cases, fetal well-being was normal beforehand, and thus currently available methods of fetal evaluation are usually not predictive. In an extreme example, Seski and Compton (1976) documented both a normal nonstress test and a normal contraction stress test done 4 hours before the onset of placental abruption that promptly killed the fetus.

PATHOLOGY. Placental abruption is initiated by hemorrhage into the decidua basalis. The decidua then splits, leaving a thin layer adherent to the myometrium. Consequently, the process in its earliest stages consists of the development of a decidual hematoma that leads to separation, compression, and the ultimate destruction of the placenta adjacent to it. In its early stage, there may be no clinical symptoms. The condition is discovered only upon examination of the freshly delivered organ, which will present on its maternal surface a circumscribed depression measuring a few centimeters in diameter, and covered by dark, clotted blood. Undoubtedly, it takes at least several minutes for these anatomical changes to materialize. Thus, a very recently separated placenta may appear no different from a normal placenta at delivery. According to Benirschke and Kaufmann (2000), the "age" of the retroplacental clot cannot be determined exactly.

In some instances, a decidual spiral artery ruptures to cause a retroplacental hematoma, which as it expands disrupts more vessels to separate more placenta. The area of separation rapidly becomes more extensive and reaches the margin of the placenta. Because the uterus is still distended by the products of conception, it is unable to sufficiently contract to compress the torn vessels that supply the placental site. The escaping blood may dissect the membranes from the uterine wall and eventually appear externally, or may be completely retained within the uterus (Figs. 25-1 and 25-2).

CONCEALED HEMORRHAGE. Retained or concealed hemorrhage is likely when:

1. There is an effusion of blood behind the placenta but its margins still remain adherent.

2. The placenta is completely separated yet the membranes retain their attachment to the uterine wall.

3. Blood gains access to the amnionic cavity after breaking through the membranes.

4. The fetal head is so closely applied to the lower uterine segment that the blood cannot make its way past it.

In most cases, however, the membranes are gradually dissected off the uterine wall, and blood sooner or later escapes.

CHRONIC PLACENTAL ABRUPTION. In some women, hemorrhage with retroplacental hematoma formation is somehow arrested completely without delivery. We have been able to document this phenomenon by labeling maternal red cells with ⁵¹chromium. This technique served to demonstrate that red blood cells concealed as clot within the uterus at delivery 3 weeks later contained no chromium and therefore were shed before.

FETAL-TO-MATERNAL HEMORRHAGE. Bleeding with placental abruption is almost always maternal. In nontraumatic placental abruption, evidence for fetomaternal hemorrhage was found in 20 percent of 78 cases; however, in all instances it was less than 10 mL (Stettler and colleagues, 1992). Significant fetal bleeding is more likely to be seen with traumatic abruption (Chap. 43, p. 1173). Pearlman and associates (1990) found fetal bleeding that averaged 12 mL in a third of women with a traumatic abruption. Stettler and colleagues (1992) reported that there was fetomaternal hemorrhage of 80 to 100 mL in 3 of 8 cases of traumatic placental abruption.

CLINICAL DIAGNOSIS. It is emphasized that the signs and symptoms with abruptio placentae can vary considerably. For example, external bleeding can be profuse, yet placental separation may not be so extensive as to compromise the fetus directly. Rarely, there may be no external bleeding but the placenta may be completely sheared off and the fetus dead as the direct consequence. In one very unusual case, a multiparous woman near term presented to the Parkland Hospital obstetrical emergency room because of nosebleed. There was no abdominal or uterine pain or tenderness and no vaginal bleeding but her fetus was dead. Her blood did not clot and the plasma fibrinogen level was 25 mg/dL. Labor was induced, and at delivery a total abruption with fresh clots was found.

Hurd and co-workers (1983), in a relatively small prospective study of abruptio placentae, identified the frequency of a variety of pertinent signs and symptoms (Table 25-4). Bleeding and abdominal pain are the most frequent findings. In 22 percent of cases, idiopathic preterm labor was considered to be the diagnosis until subsequent fetal death or distress developed. Other findings that developed were serious bleeding, back pain, uterine tenderness, frequent uterine contractions, or persistent uterine hypertonus. In older studies, ultrasound infrequently confirmed the diagnosis of abruption. For example, Sholl (1987) confirmed the clinical diagnosis sonographically in only 25 percent of women. Preliminary data from Yeo and colleagues (1999) indicated a 100-percent positive-predictive value and an 88-percent negative-predictive value in 25 women with preterm prematurely ruptured membranes and vaginal bleeding. In 48 women with vaginal bleeding and intact membranes, these corresponding numbers were 92 and 63 percent. Importantly, negative findings with ultrasound examination do not exclude placental abruption.

SHOCK. It was once held that the shock sometimes seen with placental abruption was out of proportion to the amount of hemorrhage. Supposedly, thromboplastin from decidua and placenta entered the maternal circulation and incited intravascular coagulation and other features of the amnionic fluid embolism syndrome, including hypotension. This sequence is rare, and the intensity of shock is seldom out of proportion to maternal blood loss. Pritchard and Brekken (1967) studied blood loss in 141 women with placental abruption so severe as to kill the fetus and found that it often amounted to at least half of the pregnant blood volume. Neither hypotension nor anemia is obligatory in cases of concealed hemorrhage, even when the acute hemorrhage has achieved considerable magnitude. Oliguria caused by inadequate renal perfusion but responsive to vigorous treatment of hypovolemia may also be observed in these circumstances.

DIFFERENTIAL DIAGNOSIS. In severe cases of placental abruption, the diagnosis is generally obvious. Milder and more common forms of abruption are difficult to recognize with certainty, and the diagnosis is often made by exclusion. Therefore, with vaginal bleeding complicating a viable pregnancy, it often becomes necessary to rule out placenta previa and other causes of bleeding by clinical inspection and ultrasound evaluation. It has long been taught, perhaps with some justification, that painful uterine bleeding means abruptio placentae, while painless uterine bleeding is indicative of placenta previa. Unfortunately, the differential diagnosis is not that simple. Labor accompanying placenta previa may cause pain suggestive of abruptio placentae. On the other hand, abruptio placentae may mimic normal labor, or it may cause no pain at all. The latter is more likely with a posteriorly implanted placenta.

There are neither laboratory tests nor diagnostic methods that accurately detect lesser degrees of placental separation. The cause of the vaginal bleeding at times remains obscure even after delivery. Magriples and colleagues (1999) found that thrombomodulin—an endothelial cell marker—was significantly elevated in eight women with placental abruption compared with 17 women without an abruption.

CONSUMPTIVE COAGULOPATHY. One of the most common causes of clinically significant consumptive coagulopathy in obstetrics is placental abruption. Overt hypofibrinogenemia—less than 150 mg/dL of plasma—along with elevated levels of fibrinogen-fibrin degradation products, D-dimer, and variable decreases in other coagulation factors is found in about 30 percent of women with placental abruption severe enough to kill the fetus. Such severe coagulation defects are seen less commonly in those cases in which the fetus survives. Our experience has been that serious coagulopathy, when it develops, is usually evident by the time the symptomatic woman seeks care.

The major mechanism is almost certainly the induction of coagulation intravascularly and, to a lesser degree, retroplacentally. Although an appreciable amount of fibrin is commonly deposited within the uterine cavity in cases of severe placental abruption and hypofibrinogenemia, the amounts are insufficient to account for all of the fibrinogen missing from the circulation (Pritchard and Brekken, 1967). Moreover, Bonnar and co-workers (1969) have observed, and we have confirmed, that the levels of fibrin degradation products are higher in serum from peripheral blood than in serum from blood contained in the uterine cavity. The reverse would be anticipated in the absence of significant intravascular coagulation.

An important consequence of intravascular coagulation is the activation of plasminogen to plasmin, which lyses fibrin microemboli, thereby maintaining patency of the microcirculation. In every instance of placental abruption severe enough to kill the fetus, we have identified clearly pathological levels—greater than 100 ug/mL—of fibrinogen-fibrin degradation products in maternal serum. At the outset, severe hypofibrinogenemia may or may not be accompanied by overt thrombocytopenia. After repeated blood transfusions, however, thrombocytopenia is common.

RENAL FAILURE. Acute renal failure that persists for any length of time is seen in severe forms of placental abruption. This includes those in which treatment of hypovolemia is delayed or incomplete (Chap. 47, p. 1266). Of 57 cases of acute renal failure in pregnant women described by Grunfeld and Pertuiset (1987), 23 percent were associated with placental abruption. Fortunately, reversible acute tubular necrosis accounts for three fourths of cases of renal failure (Turney and colleagues, 1989). According to Lindheimer and associates (2000), acute cortical necrosis in pregnancy is usually caused by abruptio placentae, and 7 of 19 women with this lesion indeed had a placental abruption in the report by Grunfeld and Pertuiset (1987).

Seriously impaired renal perfusion is the consequence of massive hemorrhage. Because preeclampsia frequently coexists with placental abruption, renal vasospasm is likely intensified (Hauth and Cunningham, 1999). Even when placental abruption is complicated by severe intravascular coagulation, prompt and vigorous treatment of hemorrhage with blood and crystalloid solution will often prevent clinically significant renal dysfunction. During nearly 45 years at Parkland Hospital, more than 500 cases of placental abruption so severe as to kill the fetus have received fluid replacement therapy consisting of blood and lactated Ringer solution. In only one instance has dialysis for renal failure been necessary.

For unknown reasons, proteinuria is common, especially with more severe forms of placental abruption. It usually clears soon after delivery.

COUVELAIRE UTERUS. There may be widespread extravasation of blood into the uterine musculature and beneath the uterine serosa (Fig. 25-5). This so-called uteroplacental apoplexy, first described by Couvelaire in the early 1900s, is now frequently called Couvelaire uterus. Such effusions of blood are also occasionally seen beneath the tubal serosa, in the connective tissue of the broad ligaments, and in the substance of the ovaries, as well as free in the peritoneal cavity. Its precise incidence is unknown because it can only be demonstrated conclusively at laparotomy. These myometrial hemorrhages seldom interfere with uterine contractions sufficiently to produce severe postpartum hemorrhage and are not an indication for hysterectomy.

MANAGEMENT. Treatment for placental abruption will vary depending upon gestational age and the status of the mother and fetus. With a live and mature fetus, and if vaginal delivery is not imminent, then emergent cesarean delivery is chosen by most. As discussed later in the section Hypovolemic Shock (p. 652), with massive external bleeding, intensive resuscitation with blood plus crystalloid and prompt delivery to control the hemorrhage are life saving for the mother and, it is hoped, for the fetus. If the diagnosis is uncertain and the fetus is alive but without evidence of fetal compromise, very close observation, with facilities for immediate intervention, can be practiced.

EXPECTANT MANAGEMENT IN PRETERM PREGNANCY. Delaying delivery may prove beneficial when the fetus is immature. Sholl (1987) described 72 women with pregnancies between 26 and 37 weeks who had clinically diagnosed placental abruption. About half were delivered within 3 days of admission because of progression to serious hemorrhage, fetal distress, or both. Interestingly, the cesarean rate was about 50 percent for those delivered soon after admission, as well as those in whom delivery was postponed for at least 3 days. In another study, Bond and associates (1989) expectantly managed 43 women with abruptio placentae before 35 weeks; 31 of these were given tocolytic therapy. The mean time to delivery in all 43 was about 12 days and there were no stillborns. Cesarean delivery was performed in 75 percent.

Women with evidence for very early abruption frequently develop oligohydramnios, either with or without premature membrane rupture. Elliott and associates (1998) described 24 women with abruptions who had a mean gestation of 20 weeks and who also developed oligohydraminos. They were delivered at a mean of 28 weeks.

Lack of ominous decelerations does not guarantee the safety of the intrauterine environment for any period of time. The placenta may further separate at any instant and seriously compromise or kill the fetus unless delivery is performed immediately. Some of the immediate causes of fetal distress from abruptio placentae are shown in Figure 25-6. It is important for the welfare of the distressed fetus that steps be initiated immediately to correct maternal hypovolemia, anemia, and hypoxia so as to restore and maintain the function of any placenta that is still implanted. Little can be done to favorably modify the other causes that contribute to fetal distress except to deliver the fetus.

TOCOLYSIS. Some have advocated tocolysis for preterm pregnancy complicated by suspected abruption. Hurd and associates (1983) found that abruption went unrecognized for dangerously long periods if tocolysis was initiated. Conversely, Sholl (1987) as well as Combs and co-workers (1992), provided data that tocolysis improved outcome in a highly selected group of preterm pregnancies complicated by partial abruption. Towers and co-workers (1999) administered magnesium sulfate, terbutaline, or both to 95 of 131 women with abruptio placentae diagnosed before 36 weeks. The perinatal mortality was 5 percent and did not differ from the nontreated group. They concluded that a randomized clinical trial could be safely conducted. Until then, we are of the view that clinically evident placental abruption should be considered a contraindication to tocolytic therapy.

CESAREAN DELIVERY. Rapid delivery of the fetus who is alive but in distress practically always means cesarean delivery. An electrode applied directly to the fetus may rarely provide misleading information, as in the case illustrated in Figure 25-7. At first impression at least, fetal bradycardia of 80 to 90 beats/min, with a degree of beat-to-beat variability, seemed evident. The fetus, however, was dead. There were no audible fetal heart sounds, and the maternal pulse rate was identical to that recorded through the fetal scalp electrode. Cesarean section at this time would likely have proved dangerous for the mother because she was profoundly hypovolemic and had severe consumptive coagulopathy.

VAGINAL DELIVERY. If placental separation is so severe that the fetus is dead, vaginal delivery is preferred unless hemorrhage is so brisk that it cannot be successfully managed even by vigorous blood replacement, or there are other obstetrical complications that prevent vaginal delivery. Serious coagulation defects are likely to prove especially troublesome with cesarean delivery. The abdominal and uterine incisions are prone to bleed excessively when coagulation is impaired. Hemostasis at the placental implantation site depends primarily upon myometrial contraction. Therefore, with vaginal delivery, stimulation of the myometrium pharmacologically and by uterine massage will cause these vessels to be constricted so that serious hemorrhage is avoided even though coagulation defects persist. Moreover, bleeding that does occur is shed through the vagina. An example of an indication for abdominal delivery despite documented fetal demise is now illustrated:

Although placental abruption was suspected, because rupture of a prior cesarean incision could not be excluded, repeat cesarean section was performed for a 26-week stillborn fetus. The patient had profound hypofibrinogenemia and serious bleeding was encountered from all surgical incisions. Persistent bleeding necessitated hysterectomy followed by internal iliac artery ligation. Lactated Ringer solution was given along with 17 units of blood, 8 units of plasma, and 10 units of platelets to maintain perfusion and treat the coagulopathy, which finally resolved intraoperatively.

AMNIOTOMY. Rupture of the membranes as early as possible has long been championed in the management of placental abruption. The rationale for amniotomy is that the escape of amnionic fluid might both decrease bleeding from the implantation site and reduce the entry into the maternal circulation of thromboplastin and perhaps activated coagulation factors from the retroplacental clot. There is no evidence, however, that either is accomplished by amniotomy. If the fetus is reasonably mature, rupture of the membranes may hasten delivery. If the fetus is immature, the intact sac may be more efficient in promoting cervical dilatation than will a small fetal part poorly applied to the cervix.

LABOR. With extensive placental abruption, the uterus will likely be persistently hypertonic. The baseline intra-amnionic pressure may be 50 mm Hg or higher, with rhythmic increases up to 75 to 100 mm Hg. Because of persistent hypertonus, it may be difficult at times to determine by palpation if the uterus is contracting and relaxing to any degree (Fig. 25-8).

OXYTOCIN. Although hypertonicity characterizes myometrial function in most cases of severe placental abruption, if no rhythmic uterine contractions are superimposed, then oxytocin is given in standard doses. Uterine stimulation to effect vaginal delivery provides benefits that override the risks. The use of oxytocin has been challenged on the basis that it might enhance the escape of thromboplastin into the maternal circulation and thereby initiate or enhance consumptive coagulopathy or amnionic fluid embolism syndrome. There is no evidence to support this fear (Clark and colleagues, 1995; Pritchard and Brekken, 1967).

TIMING OF DELIVERY AFTER SEVERE PLACENTAL ABRUPTION. When the fetus is dead or previable, there is no evidence that establishing an arbitrary time limit for delivery is necessary. Experiences at both the University of Virginia and Parkland Hospitals indicate that the maternal outcome depends upon the diligence with which adequate fluid and blood replacement therapy is pursued, rather than upon the interval to delivery (Brame and associates, 1968; Pritchard and Brekken, 1967). At the University of Virginia Hospital, women with severe placental abruption who were transfused for 18 hours or more before delivery, experienced complications that were neither more numerous nor greater in severity than did the group in which delivery was accomplished sooner. Our observations are similar; Figure 25-9 summarizes serial findings from one of the most severe cases in terms of the prolonged interval between the onset of symptoms and delivery and the necessity of transfusing a large volume of blood.

PLACENTA PREVIA

DEFINITION. In placenta previa, the placenta is located over or very near the internal os. Four degrees of this abnormality have been recognized:

1. Total placenta previa. The internal cervical os is covered completely by placenta (Fig. 25-10).

2. Partial placenta previa. The internal os is partially covered by placenta (Figs. 25-1 and 25-10).

3. Marginal placenta previa. The edge of the placenta is at the margin of the internal os.

4. Low-lying placenta. The placenta is implanted in the lower uterine segment such that the placenta edge actually does not reach the internal os but is in close proximity to it.

Another condition, termed vasa previa, is where the fetal vessels course through membranes and present at the cervical os. This is an uncommon cause of antepartum hemorrhage and is associated with a high rate of fetal death. Prenatal diagnosis by ultrasonography improves perinatal salvage (Lee and colleagues, 2000). It is discussed in detail in Chapter 32.

The degree of placenta previa will depend in large measure on the cervical dilatation at the time of examination. For example, a low-lying placenta at 2 cm dilatation may become a partial placenta previa at 8 cm dilatation because the dilating cervix has uncovered placenta. Conversely, a placenta previa that appears to be total before cervical dilatation may become partial at 4 cm dilatation because the cervix dilates beyond the edge of the placenta (Fig. 25-11). Digital palpation to try to ascertain these changing relations between the edge of the placenta and the internal os as the cervix dilates can incite severe hemorrhage!

In both the total and partial placenta previa, a certain degree of spontaneous placental separation is an inevitable consequence of the formation of the lower uterine segment and cervical dilatation. Such separation is associated with hemorrhage from blood vessels so disrupted.

INCIDENCE. Iyasu and co-workers (1993), in an analysis of the National Hospital Discharge Survey from 1979 to 1987, found that placenta previa complicated 0.5 percent (1 in 200) deliveries. At Prentice Women's Hospital, Frederiksen and colleagues (1999) reported that 0.55 percent (1 in 180) of nearly 93,500 deliveries were complicated by previa. Crane and associates (1999) found the incidence to be 0.33 percent (1 in 300) in almost 93,000 deliveries in the province of Nova Scotia. At Parkland Hospital, the incidence was 0.26 percent (1 in 390) for more than 169,000 deliveries over 12 years.

These statistics are remarkably similar considering the lack of precision in definition and identification for reasons already discussed. A question difficult to answer is whether painless bleeding from focal separation of a placenta implanted in the lower uterine segment but away from a partially dilated cervical os should be classified as placenta previa or placental abruption. Obviously, it is both.

ETIOLOGY. Advancing maternal age increases the risk of placenta previa. As shown in Figure 25-4, in over 169,000 deliveries at Parkland Hospital from 1988 through 1999, the incidence of previa increased significantly with each age group. At the extremes, it is 1 in 1500 for women 19 or less and for women over 35 it is 1 in 100. Frederiksen and colleagues (1999) reported that the incidence of previa increased from 0.3 percent in 1976 to 0.7 percent in 1997. They attributed this to a shift to an older obstetrical population.

Multiparity is associated with previa. In a study of 314 women who were para 5 or greater, Babinszki and collaborators (1999) reported that the incidence of previa of 2.2 percent was increased significantly compared with women of lower parity. In the 169,000-plus women at Parkland Hospital the incidence was 1 in 175 in women para 3 or greater.

Prior cesarean delivery increases the likelihood of placenta previa. Nielsen and colleagues (1989) found a fivefold increased incidence of placenta previa in Swedish women with a prior cesarean delivery. At Parkland, the incidence was increased twofold from 1 in 400 to 1 in 200 with at least one prior cesarean section. Miller and associates (1996) cited a threefold increase of previa in women with prior cesarean delivery in over 150,000 deliveries at Los Angeles County Women's Hospital. The incidence increased with the number of previous cesarean deliveries—it was 1.9 percent with two prior cesareans and 4.1 percent with three or more. Certainly, a prior cesarean incision with a previa increases the incidence of hysterectomy. Frederiksen and co-workers (1999) reported a 25-percent hysterectomy rate in women with repeat cesarean for a previa compared with only 6 percent of those undergoing primary cesarean for placenta previa.

Williams and colleagues (1991b) found the relative risk of placenta previa to be increased twofold related to smoking. They theorized that carbon monoxide hypoxemia caused compensatory placental hypertrophy. These findings were confirmed by Handler and colleagues (1994). Perhaps related, defective decidual vascularization, the possible result of inflammatory or atrophic changes, has been implicated in the development of previa.

PLACENTA ACCRETA, INCRETA, AND PERCRETA. Placenta previa may be associated with placenta accreta or one of its more advanced forms, placenta increta or percreta. Such abnormally firm attachment of the placenta might be anticipated because of poorly developed decidua in the lower uterine segment. Almost 7 percent of 514 cases of previa reported by Frederiksen and collaborators (1999) had an associated abnormal placental attachment. Biswas and co-workers (1999) performed placental bed biopsies at cesarean delivery in 50 women with previas and 50 control women. While about half of specimens from previas showed myometrial spiral arterioles with trophoblastic giant-cell infiltration, only 20 percent from those normally implanted had these changes.

CLINICAL FINDINGS. The most characteristic event in placenta previa is painless hemorrhage, which usually does not appear until near the end of the second trimester or after. Some abortions, however, may result from such an abnormal location of the developing placenta. Frequently, bleeding from placenta previa has its onset without warning, presenting without pain in a woman who has had an uneventful prenatal course. Fortunately, the initial bleeding is rarely so profuse as to prove fatal. Usually it ceases spontaneously, only to recur. In some cases, particularly those with a placenta implanted near but not over the cervical os, bleeding does not appear until the onset of labor, when it may vary from slight to profuse hemorrhage and may clinically mimic placental abruption.

The cause of hemorrhage is reemphasized. When the placenta is located over the internal os, the formation of the lower uterine segment and the dilatation of the internal os result inevitably in tearing of placental attachments. The bleeding is augmented by the inability of the myometrial fibers of the lower uterine segment to contract and thereby constrict the torn vessels.

Hemorrhage from the placental implantation site in the lower uterine segment may continue after delivery of the placenta, because the lower uterine segment is more prone to contract poorly than is the uterine body. Bleeding may also result from lacerations in the friable cervix and lower uterine segment, especially following manual removal of a somewhat adherent placenta.

COAGULATION DEFECTS. In our experiences, coagulopathy is rare with placenta previa, even when extensive separation from the implantation site has occurred. Wing and colleagues (1996b) studied 87 women with antepartum bleeding from placenta previa and found no evidence for coagulopathy. Presumably thromboplastin that incites intravascular coagulation that commonly characterizes abruptio placentae readily escapes through the cervical canal rather than being forced into the maternal circulation.

DIAGNOSIS. In women with uterine bleeding during the latter half of pregnancy, placenta previa or abruptio placentae should always be suspected. The possibility of placenta previa should not be dismissed until appropriate evaluation, including sonography, has clearly proved its absence. The diagnosis of placenta previa can seldom be established firmly by clinical examination unless a finger is passed through the cervix and the placenta is palpated. Such examination of the cervix is never permissible unless the woman is in an operating room with all the preparations for immediate cesarean delivery, because even the gentlest examination of this sort can cause torrential hemorrhage. Furthermore, such an examination should not be made unless delivery is planned, for it may cause bleeding of such a degree that immediate delivery becomes necessary even though the fetus is immature. Such a "double set-up" examination is rarely necessary as placental location can almost always be obtained by sonography.

LOCALIZATION BY SONOGRAPHY. The simplest, most precise, and safest method of placental localization is provided by transabdominal sonography, which is used to locate the placenta with considerable accuracy (Figs. 25-12 and 25-13). According to Laing (1996), the average accuracy is about 96 percent, and rates as high as 98 percent have been obtained. False-positive results are often a result of bladder distention. Therefore, ultrasonic scans in apparently positive cases should be repeated after emptying the bladder. An uncommon source of error has been identification of abundant placenta implanted in the uterine fundus but failure to appreciate that the placenta was large and extended downward all the way to the internal os of the cervix.

The use of transvaginal ultrasonography has substantively improved diagnostic accuracy of placenta previa. Farine and associates (1988) were able to visualize the internal cervical os in all cases with the transvaginal technique, in contrast to only 70 percent using transabdominal equipment. An example is shown in Figure 25-14. Leerentveld and colleagues (1990) studied 100 women suspected of having placenta previa. They reported a 93-percent positive-predictive value and 98-percent negative-predictive value for transvaginal ultrasonography. Tan and co-workers (1995) reported less accuracy with the technique. In studies comparing abdominal ultrasound with transvaginal imaging, Smith (1997) and Taipale (1998) and their colleagues found the transvaginal technique to be superior. Most now agree that confirmatory transvaginal imaging is indicated if the placenta is low lying or appears to be covering the cervical os by transabdominal sonography.

Hertzberg and associates (1992) demonstrated that transperineal sonography allowed visualization of the internal os in all 164 cases examined because transabdominal sonography disclosed a previa or was inconclusive. Placenta previa was correctly excluded in 154 women, and in 10 in whom it was diagnosed sonographically, nine had a previa confirmed at delivery. The positive-predictive value was 90 percent and the negative-predictive value was 100 percent.

MAGNETIC RESONANCE IMAGING. A number of investigators have used magnetic resonance imaging to visualize placental abnormalities, including placenta previa. Kay and Spritzer (1991) discussed the many positive attributes of such technology (Fig. 25-15). It is unlikely that this will replace ultrasonic scanning for routine evaluation in the near future.

PLACENTAL "MIGRATION." Since the report by King (1973), the apparent peripatetic nature of the placenta has been well established. In earlier studies, "low-lying placentas" were evaluated and McClure and Dornal (1990) found these in 25 percent of 1490 routine scans at 18 weeks. At delivery, only seven of these 385 low-lying placentas persisted. Sanderson and Milton (1991) found that only 12 percent of placentas were "low lying" in 4300 women at 18 to 20 weeks. Of those not covering the interal os, previa did not persist and hemorrhage was not encountered. Conversely, of those covering the os at midpregnancy, about 40 percent persisted as a previa.

Thus, placentas that lie close to the internal os, but not over it, during the second trimester, or even early in the third trimester, are very unlikely to persist as previas by term. These data were amplified by Taipale and co-workers (1998), who found that 57 of 3696 (1.5 percent) unselected women had a placenta previa at 18 to 23 weeks. Only 20 percent of those with the placental edge extending less than 15 mm from the os had a previa at delivery. If the placental edge extended to 25 mm or more, however, 40 percent had a previa.

The low frequency with which placenta previa persists when it has been identified sonographically before 30 weeks is shown in Table 25-5. It is apparent from these data that in the absence of any other abnormality, sonography need not be frequently repeated simply to follow placental position, and restriction of activity need not be practiced unless the previa persists beyond 30 weeks, or becomes clinically apparent before that time.

The mechanism of apparent placental movement is not completely understood. The term migration is clearly a misnomer, however, as invasion of chorionic villi into the decidua on either side of the cervical os will persist. The apparent movement of the low-lying placenta relative to the internal os probably results from inability to precisely define this relationship in a three-dimensional manner using two-dimensional sonography in early pregnancy. This difficulty is coupled with differential growth of lower and upper myometrial segments as pregnancy progresses. Thus, those placentas that "migrate" most likely never had actual circumferential villus invasion that reached the internal cervical os in the first place.

MANAGEMENT. Women with a placenta previa may be considered as follows:

1. Those in whom the fetus is preterm but there is no indication for delivery.

2. Those in whom the fetus is reasonably mature.

3. Those in labor.

4. Those in whom hemorrhage is so severe as to mandate delivery despite fetal immaturity.

Management with a preterm fetus, but with no active bleeding, consists of close observation. In some cases prolonged hospitalization may be ideal; however, the woman is usually discharged after bleeding has ceased and the fetus is judged to be healthy. The women and her family must fully appreciate the problems of placenta previa and be prepared to transport her to the hospital immediately. In properly selected patients, there appears to be no benefit to inpatient versus outpatient management of placenta previa (Mouer, 1994). Drost and Keil (1994) demonstrated a 50 percent reduction in hospital days, a 50 percent reduction in maternal cost, and a 40 percent reduction in cost for mother-infant pairs, with no differences in maternal or fetal morbidity with outpatient compared with inpatient management. Wing and colleagues (1996a) reported preliminary results from their randomized clinical trial of inpatient versus home management of 53 women with bleeding from a previa at 24 to 36 weeks. Maternal and perinatal morbidity was similar in each group, but home management saved $15,000 per case. Importantly, 33 (62 percent) of these 53 women had recurrent bleeding and in 28 it required expeditious cesarean delivery.

DELIVERY. Cesarean delivery is necessary in practically all cases of placenta previa. In most cases a transverse uterine incision is made. Because fetal bleeding may result from an incision into an anterior placenta, a vertical incision is sometimes recommended in these circumstances. Even when the incision extends through the placenta, however, maternal or fetal outcome is rarely compromised.

Because of the poorly contractile nature of the lower uterine segment, there may be uncontrollable hemorrhage following placental removal. This can occur even without histological confirmation of placenta accreta. Under these circumstances, management appropriate for placenta accreta is indicated (see p. 632). When placenta previa is complicated by degrees of placenta accreta that render control of bleeding from the placental bed difficult by conservative means, other methods of hemostasis are necessary. Oversewing the implantation site with 0-chromic sutures may provide hemostasis. In some cases, bilateral uterine artery ligation is helpful, and in others, bleeding ceases with internal iliac artery ligation. Cho and colleagues (1991) have described placing circular interrupted 0-chromic sutures around the lower segment, above and below the transverse incision, which controlled hemorrhage in all eight women in whom this was employed. Druzin (1989) described four cases in which the lower uterine segment was tightly packed with gauze that successfully arrested hemorrhage. The pack was removed transvaginally 12 hours later.

If these conservative methods fail, and bleeding is brisk, then hysterectomy is necessary. In some cases, uterine or internal iliac artery ligation as described on page 652 may provide hemostasis. Pelvic artery embolization has gained acceptance also (Hansch and colleagues, 1999; Pelage and associates, 1999). For women whose placenta previa is implanted anteriorly in the site of a prior cesarean section incision, there is an increased likelihood of associated placenta accreta and need for hysterectomy (see p. 632).

PROGNOSIS. A marked reduction in maternal mortality from placenta previa has been achieved, a trend that began in 1927 when Bill advocated adequate transfusion and cesarean delivery. Since 1945, when Macafee and Johnson independently suggested expectant therapy for patients remote from term, a similar trend has been evident in perinatal loss. Although half of women are near term when bleeding first develops, preterm delivery still poses a formidable problem for the remainder, because not all women with placenta previa and a preterm fetus can be treated expectantly.

PERINATAL MORBIDITY AND MORTALITY. Preterm delivery is a major cause of perinatal death even though expectant management of placenta previa is practiced. In their study of almost 93,000 deliveries, Crane and co-investigators (1999) reported a preterm delivery rate of 47 percent. Mortality from complications of preterm birth, however, was not increased when compared with infants of similar gestational age born to women without a previa. Although suggested earlier by some investigators that congenital malformations are increased with a previa, Crane and co-workers (1999) were the first to confirm this and to control for maternal age. For reasons that are unclear, fetal anomalies were increased 2.5-fold.

It is unclear if there is associated fetal growth restriction with a previa. Brar and colleagues (1988) reported that the incidence was nearly 20 percent. Conversely, Crane and co-workers (1999) found no increased incidence after controlling for gestational age. Wolf and associates (1991), in a case-control study of 179 women with placenta previa, found the incidence of growth restriction to be 5 percent for both groups.

POSTPARTUM HEMORRHAGE

Hemorrhage following delivery is the consequence of excessive bleeding from the placental implantation site, trauma to the genital tract and adjacent structures, or both (Table 25-6). Thus, postpartum hemorrhage is a description of an event, and not a diagnosis. In the United Kingdom, half of maternal deaths from hemorrhage are due to postpartum events (Bonnar, 2000). When excess bleeding is encountered, a specific etiology should be sought. Uterine atony, degrees of retained placenta—including placenta accreta and its variants, and genital tract lacerations account for most cases of postpartum hemorrhage. In the past 20 years, placenta accreta has overtaken uterine atony as the most common cause of postpartum hemorrhage of sufficient severity to mandate hysterectomy (Chestnut and colleagues, 1985; Clark and associates, 1984; Zelop and co-workers, 1993).

DEFINITION. Traditionally postpartum hemorrhage has been defined as the loss of 500 mL or more of blood after completion of the third stage of labor. Nonetheless, nearly a half of all women who are delivered vaginally shed that amount of blood or more, when measured quantitatively (Fig. 25-16). This compares with 1000 mL blood loss for cesarean section, 1400 mL for elective cesarean hysterectomy, and 3000 to 3500 mL for emergency cesarean hysterectomy (Chestnut and associates, 1985; Clark and colleagues, 1984). The woman with normal pregnancy-induced hypervolemia usually increases her blood volume by 30 to 60 percent, which for an average-sized woman amounts to 1 to 2 L (Pritchard, 1965). Consequently, she will tolerate, without any remarkable decrease in postpartum hematocrit, blood loss at delivery that approaches the volume of blood she added during pregnancy. In one study, the mean postpartum hematocrit decline ranged from 2.6 to 4.3 volume percent; a third of women had no decline or had an actual increase (Combs and colleagues, 1991b). Women undergoing cesarean delivery had a mean drop in hematocrit of 4.2 volume percent, but 20 percent had no decline (Combs and co-workers, 1991a).

Therefore, blood loss somewhat in excess of 500 mL by accurate measurement is not necessarily an abnormal event for vaginal delivery. Pritchard and associates (1962) found that about 5 percent of women delivering vaginally lost more than 1000 mL of blood. They also observed that estimated blood loss is commonly only about half the actual loss. Based on an estimated blood loss greater than 500 mL, postpartum hemorrhage has been found in about 5 percent of deliveries. An estimated blood loss in excess of 500 mL in many institutions, therefore, should call attention to mothers who are bleeding excessively and warn the physician that dangerous hemorrhage is imminent. Hemorrhage after the first 24 hours is designated late postpartum hemorrhage and is discussed in Chapter 17 (p. 406).

HEMOSTASIS AT THE PLACENTAL SITE. Near term, it is estimated that approximately 600 mL/min of blood flows through the intervillous space. With separation of the placenta, the many uterine arteries and veins that carry blood to and from the placenta are severed abruptly. Elsewhere in the body, hemostasis in the absence of surgical ligation depends upon intrinsic vasospasm and formation of blood clot locally. At the placental implantation site, most important for achieving hemostasis are contraction and retraction of the myometrium to compress the vessels and obliterate their lumens. Adherent pieces of placenta or large blood clots will prevent effective contraction and retraction of the myometrium and thereby impair hemostasis at the implantation site. Fatal postpartum hemorrhage can occur from a hypotonic uterus while the maternal blood coagulation mechanism is quite normal. Conversely, if the myometrium at and adjacent to the denuded implantation site contracts and retracts vigorously, fatal hemorrhage from the placental implantation site is unlikely even though the coagulation mechanism is severely impaired.

PROLONGED THIRD STAGE. Occasionally, the placenta does not separate promptly. A question to which there is still no definite answer concerns the length of time that should elapse in the absence of bleeding before the placenta is removed manually. Management is discussed in Chapter 13 (p. 321). Obstetrical tradition has set somewhat arbitrary limits on third-stage duration in attempts to define abnormally retained placenta, and thus reduce blood loss due to excessively prolonged placental separation. Combs and Laros (1991) studied 12,275 singleton vaginal deliveries and reported the median third-stage duration to be 6 minutes, and 3.3 percent were more than 30 minutes. Several measures of hemorrhage, including curettage or transfusion, increased with third stages nearing 30 minutes or longer.

UTERINE ATONY

Failure of the uterus to contract properly following delivery is a common cause of obstetrical hemorrhage. In many cases, postpartum hemorrhage can be predicted well in advance of delivery (Table 25-6). Examples in which trauma may lead to postpartum hemorrhage include delivery of a large infant, midforceps delivery, forceps rotation, any intrauterine manipulation, and perhaps vaginal delivery after cesarean section or other uterine incisions. Uterine atony causing hemorrhage can be anticipated whenever excessive concentrations of halogenated anesthetic agents are used that will relax the uterus (Gilstrap and colleagues, 1987). The overdistended uterus is very likely to be hypotonic after delivery. Thus, the woman with a large fetus, multiple fetuses, or hydramnios is prone to hemorrhage from uterine atony. Blood loss with delivery of twins, for example, averages nearly 1000 mL and may be much greater (Pritchard, 1965). The woman whose labor is characterized by uterine activity that is either remarkably vigorous or barely effective is also likely to bleed excessively from uterine atony after delivery.

Similarly, labor either initiated or augmented with oxytocin is more likely to be followed by postdelivery uterine atony and hemorrhage. The woman of high parity may be at increased risk for uterine atony. Fuchs and colleagues (1985) described the outcomes of nearly 5800 women para 7 or greater. They reported that the 2.7 percent incidence of postpartum hemorrhage in these women was increased fourfold compared with the general obstetrical population. Babinszki and colleagues (1999) reported the incidence of postpartum hemorrhage to be 0.3 percent in women of low parity, but it was 1.9 percent in those para 4 or greater.

Another risk is if the woman has previously suffered postpartum hemorrhage. Finally, mismanagement of the third stage of labor involves an attempt to hasten delivery of the placenta short of manual removal. Constant kneading and squeezing of the uterus that already is contracted likely will impede the physiological mechanism of placental detachment, causing incomplete placental separation and increased blood loss.

CLINICAL CHARACTERISTICS. Postpartum hemorrhage before placental delivery is called third-stage hemorrhage. Contrary to general opinion, whether bleeding begins before or after placental delivery, or at both times, there may be no sudden massive hemorrhage but rather steady bleeding that at any given instant appears to be moderate, but persists until serious hypovolemia develops. Especially with hemorrhage after placental delivery, the constant seepage may lead to enormous blood loss.

The effects of hemorrhage depend to a considerable degree upon the nonpregnant blood volume, magnitude of pregnancy-induced hypervolemia, and degree of anemia at the time of delivery. A treacherous feature of postpartum hemorrhage is the failure of the pulse and blood pressure to undergo more than moderate alterations until large amounts of blood have been lost. The normotensive woman may actually become somewhat hypertensive in response to hemorrhage, at least initially. Moreover, the already hypertensive woman may be interpreted to be normotensive although remarkably hypovolemic. Tragically, the hypovolemia may not be recognized until very late.

As emphasized in Chapter 24, the woman with severe preeclampsia has usually lost her pregnancy-induced hypervolemia. Thus, she frequently is very sensitive or even intolerant of what may be considered normal blood loss. Therefore, when excessive hemorrhage is even suspected in the woman with severe pregnancy-induced hypertension, efforts should be made immediately to identify those clinical and laboratory findings that would prompt vigorous crystalloid and blood replacement.

In instances in which the fundus has not been adequately monitored after delivery, the blood may not escape vaginally but instead may collect within the uterus. The uterine cavity may thus become distended by 1000 mL or more of blood while an inattentive attendant fails to identify the large uterus or, having done so, erroneously massages a roll of abdominal fat. The care of the postpartum uterus, therefore, must not be left to an inexperienced person.

DIAGNOSIS. Except possibly when intrauterine and intravaginal accumulation of blood are not recognized, or in some instances of uterine rupture with intraperitoneal bleeding, the diagnosis of postpartum hemorrhage should be obvious. The differentiation between bleeding from uterine atony and from lacerations is tentatively made on the condition of the uterus. If bleeding persists despite a firm, well-contracted uterus, the cause of the hemorrhage most probably is from lacerations. Bright red blood also suggests lacerations. To ascertain the role of lacerations as a cause of bleeding, careful inspection of the vagina, cervix, and uterus is essential.

Sometimes bleeding may be caused by both atony and trauma, especially after major operative delivery. In general, inspection of the cervix and vagina should be performed after every delivery to identify hemorrhage from lacerations. Anesthesia should be adequate to prevent discomfort during such an examination. Examination of the uterine cavity, the cervix, and all of the vagina is essential after breech extraction, after internal podalic version, and following vaginal delivery in a woman who previously underwent cesarean section. The same is true when unusual bleeding is identified during the second stage of labor.

SHEEHAN SYNDROME. Severe intrapartum or early postpartum hemorrhage is on rare occasions followed by pituitary failure. In the classical case of Sheehan syndrome, this is characterized by failure of lactation, amenorrhea, breast atrophy, loss of pubic and axillary hair, hypothyroidism, and adrenal cortical insufficiency (Chap. 50, p. 1354). The exact pathogenesis is not well understood, because such endocrine abnormalities do not develop in most women who hemorrhage severely. In some but not all instances of Sheehan syndrome, varying degrees of anterior pituitary necrosis with impaired secretion of one or more trophic hormones account for endocrine abnormalities. The anterior pituitary of some women who develop hypopituitarism after puerperal hemorrhage does respond to various releasing hormones, which at the least implies impaired hypothalamic function. Moreover, Whitehead (1963) identified specific atrophic changes in hypothalamic nuclei histologically in some cases. Lactation after delivery usually, but not always, excludes extensive pituitary necrosis. In some women, failure to lactate may not be followed until many years later by other symptoms of pituitary insufficiency. In the series reported by Ammini and Mathur (1994), the average duration of onset of symptoms was 5 years.

The incidence of Sheehan syndrome was originally estimated to be 1 per 10,000 deliveries (Sheehan and Murdoch, 1938). It appears to be even more rare today in the United States. Application of tests of hypothalamic and pituitary function now available should identify milder forms of the syndrome and define their prevalence (Grimes and Brooks, 1980). Bakiri and colleagues (1991) used computed tomography to study 54 women with documented Sheehan syndrome. In all of these, the appearance of the pituitary was abnormal; the sella turcica was either totally or partially empty.

MANAGEMENT OF THIRD-STAGE BLEEDING. Some bleeding is inevitable during the third stage as the result of transient partial separation of the placenta. As the placenta separates, the blood from the implantation site may escape into the vagina immediately (Duncan mechanism) or it may be concealed behind the placenta and membranes (Schultze mechanism) until the placenta is delivered.

In the presence of any external hemorrhage during the third stage, the uterus should be massaged if it is not contracted firmly. If the signs of placental separation have appeared, expression of the placenta should be attempted by manual fundal pressure as described in Chapter 13 (p. 321). Descent of the placenta is indicated by the cord becoming slack. If bleeding continues, manual removal of the placenta is mandatory.

TECHNIQUE OF MANUAL REMOVAL. Adequate analgesia or anesthesia is mandatory. Aseptic surgical technique should be employed. After grasping the fundus through the abdominal wall with one hand, the other hand is introduced into the vagina and passed into the uterus, along the umbilical cord. As soon as the placenta is reached, its margin is located and the ulnar border of the hand insinuated between it and the uterine wall (Fig. 25-17). Then with the back of the hand in contact with the uterus, the placenta is peeled off its uterine attachment by a motion similar to that employed in separating the leaves of a book. After its complete separation, the placenta should be grasped with the entire hand, which is then gradually withdrawn. Membranes are removed at the same time by carefully teasing them from the decidua, using ring forceps to grasp them as necessary. Some clinicians prefer to wipe out the uterine cavity with a sponge. If this is done, it is imperative that a sponge not be left in the uterus or vagina.

MANAGEMENT AFTER DELIVERY OF PLACENTA. The fundus should always be palpated following placental delivery to make certain that the uterus is well contracted. If it is not firm, vigorous fundal massage is indicated. Most often, 20 U of oxytocin in 1000 mL of lactated Ringer or normal saline proves effective when administered intravenously at approximately 10 mL/min (200 mU of oxytocin per minute) simultaneously with effective uterine massage. Oxytocin should never be given as an undiluted bolus dose as serious hypotension or cardiac arrhythmias may follow (Chap. 13, p. 323).

ERGOT DERIVATIVES. If oxytocin given by rapid infusion does not prove effective, some administer methylergonovine, 0.2 mg intramuscularly or intravenously. This may stimulate the uterus to contract sufficiently to control hemorrhage. Any superior therapeutic effects of ergot derivatives over oxytocin are speculative, and if intravenously administered they may cause dangerous hypertension, especially in the woman with preeclampsia.

PROSTAGLANDINS. The 15-methyl derivative of prostaglandin F_2a (carboprost tromethamine) was approved in the mid-1980s by the Food and Drug Administration for treatment of uterine atony. The initial recommended dose is 250 ug (0.25 mg) given intramuscularly, and this is repeated if necessary at 15- to 90-minute intervals up to a maximum of eight doses. Oleen and Mariano (1990) studied use of carboprost for postpartum hemorrhage at 12 cooperating obstetrical units. Arrest of bleeding was considered successful in 208 of 237 (88 percent) women treated. An additional 17 women required other oxytocics for control of hemorrhage. The remaining 12 women in whom drug treatment failed required surgical intervention.

Carboprost is associated with side effects in about 20 percent of women (Oleen and Mariano, 1990). In descending order of frequency, these include diarrhea, hypertension, vomiting, fever, flushing, and tachycardia. We have encountered serious hypertension in a few women so treated. In addition, Hankins and colleagues (1988) observed that intramuscular carboprost was followed by arterial oxygen desaturation that averaged 10 percent and developed within 15 minutes. They concluded that this was due to pulmonary airway and vascular constriction.

Rectally administered prostaglandin E₂ 20-mg suppositories have been used for uterine atony, but not studied in clinical trials. O'Brien and colleagues (1998) reported that misoprostol, 1000 ug given rectally, was effective in 14 women unresponsive to usual oxytocics.

BLEEDING UNRESPONSIVE TO OXYTOCICS. Continued bleeding after multiple oxytocic administrations may be from unrecognized genital tract lacerations, including in some cases uterine rupture. Thus, if bleeding persists, no time should be lost in haphazard efforts to control hemorrhage, but the following management should be initiated immediately:

1. Employ bimanual uterine compression (Fig. 25-18). The technique of consists simply of massage of the posterior aspect of the uterus with the abdominal hand and massage through the vagina of the anterior uterine aspect with the other fist. This procedure will control most hemorrhage.

2. Obtain help!

3. Begin blood transfusions. The blood group of every obstetrical patient should be known, if possible, before labor, and an indirect Coombs test done to detect erythrocyte antibodies. If the latter is negative, then cross-matching of blood is not necessary (see p. 655). In an extreme emergency, type O D-negative "universal donor" blood is given.

4. Explore the uterine cavity manually for retained placental fragments or lacerations.

5. Thoroughly inspect the cervix and vagina after adequate exposure.

6. Add a second large-bore intravenous catheter so that crystalloid with oxytocin is continued at the same time as blood is given.

7. A Foley catheter is inserted to monitor urine output, which is a good measure of renal perfusion.

Resuscitation is then carried out as described subsequently (see p. 652). Blood transfusion should be considered in any case of postpartum hemorrhage in which abdominal uterine massage and oxytocic agents fail to control the bleeding. With transfusion and simultaneous manual uterine compression and intravenous oxytocin, additional measures are rarely required. Intractable atony may mandate hysterectomy as a lifesaving measure (Chap. 23, p. 553). Alternatively, uterine artery ligation, internal iliac artery ligation, or angiographic embolization as described on page 647 may prove successful.

HEMORRHAGE FROM RETAINED PLACENTAL FRAGMENTS

Immediate postpartum hemorrhage is seldom caused by retained small placental fragments, but a remaining piece of placenta is a common cause of bleeding late in the puerperium. Inspection of the placenta after delivery must be routine. If a portion of placenta is missing, the uterus should be explored and the fragment removed, particularly with continuing postpartum bleeding. Retention of a succenturiate lobe (see Fig. 5-15) is an occasional cause of postpartum hemorrhage. The late bleeding that may result from a placental polyp is discussed in Chapter 17 (p. 406).

PLACENTA ACCRETA, INCRETA, AND PERCRETA

In most instances, the placenta separates spontaneously from its implantation site during the first few minutes after delivery of the infant. The precise reason for delay in detachment beyond this time is not always obvious, but quite often it seems to be due to inadequate uterine contraction. Very infrequently, the placenta is unusually adherent to the implantation site, with scanty or absent decidua, so that the physiological line of cleavage through the decidual spongy layer is lacking. As a consequence, one or more cotyledons are firmly bound to the defective decidua basalis or even to the myometrium. When the placenta is densely anchored in this fashion, the condition is called placenta accreta.

DEFINITIONS. The term placenta accreta is used to describe any placental implantation in which there is abnormally firm adherence to the uterine wall. As the consequence of partial or total absence of the decidua basalis and imperfect development of the fibrinoid layer (Nitabuch layer), placental villi are attached to the myometrium in placenta accreta, actually invade the myometrium in placenta increta, or penetrate through the myometrium in placenta percreta (Figs. 25-19 and 25-20). The abnormal adherence may involve all of the cotyledons (total placenta accreta), a few to several cotyledons (partial placenta accreta), or a single cotyledon (focal placenta accreta). According to Benirschke and Kaufmann (2000), histological diagnosis of accreta cannot be made from the placenta alone and the entire uterus or curettings with myometrium are necessary.

SIGNIFICANCE. An abnormally adherent placenta, although an uncommon condition, assumes considerable significance clinically because of morbidity and, at times, mortality from severe hemorrhage, uterine perforation, and infection. The incidence of placenta accreta, increta, and percreta have increased because of the increased cesarean delivery rate. In earlier reports, Breen and associates (1977) reviewed studies published since 1891 and found an average incidence of about 1 in 7000. Read and co-workers (1980) reported an incidence of about 1 per 2500 deliveries. Zelop and colleagues (1993) reported that abnormally adherent placentation caused 65 percent of cases of intractable postpartum hemorrhage requiring emergency peripartum hysterectomy at Brigham and Women's Hospital. Zaki and associates (1998) found an incidence of 1 in 1900 from 1990 to 1996 during which time there were over 23,000 deliveries at their hospital.

ETIOLOGICAL FACTORS. Abnormal placental adherence is found when decidual formation is defective. Associated conditions include implantation in the lower uterine segment; over a previous cesarean section scar or other previous uterine incisions; or after uterine curettage. In his review of 622 cases collected between 1945 and 1969, Fox (1972) noted the following characteristics:

1. Placenta previa was identified in a third of affected pregnancies.

2. A fourth of the women had been previously delivered by cesarean section.

3. Nearly a fourth had previously undergone curettage.

4. A fourth were gravida 6 or more.

Zaki and associates (1998) found that 10 percent of 112 consecutive cases of placenta previa had associated accreta. Hardardottir and colleagues (1996) observed that almost half of placentas in women with a prior cesarean delivery had adherent myometrial fibers detected microscopically.

Other risk factors for placenta accreta were analyzed by Hung and co-workers (1999) in their study of over 9300 women screened for Down syndrome at 14 to 22 weeks. Compared with the normal population, they found a 54-fold increased risk for accreta with placenta previa; an 8.3-fold risk when maternal serum alpha-fetoprotein levels exceeded 2.5 multiples of the median (MOM); or a 3.9-fold risk when maternal free b-hCG levels were greater than 2.5 MOM; and a 3.2-fold risk with maternal age of 35 or older.

CLINICAL COURSE AND DIAGNOSIS. Early in pregnancy, the maternal serum alpha-fetoprotein level may be increased (Chap. 37, p. 983). Antepartum hemorrhage is common, but in the great majority of cases, bleeding before delivery is the consequence of coexisting placenta previa. Myometrial invasion by placental villi at the site of a previous cesarean section scar may lead to uterine rupture before labor (Berchuck and Sokol, 1983). We have seen this as early as 12 weeks in a woman explored for an ectopic pregnancy. Archer and Furlong (1987) described massive hemoperitoneum caused by placenta percreta at 21 weeks. In women whose pregnancies go to term, however, labor will most likely be normal in the absence of an associated placenta previa or an involved uterine scar.

The possibility exists that placenta increta might be diagnosed antepartum. Cox and associates (1988) described a case of placenta previa in which they also were able to identify placenta increta ultrasonically from the lack of the usual subplacental sonolucent space. They hypothesize that the presence of this normal subplacental sonolucent area represents the decidua basalis and the underlying myometrial tissue. The absence of this sonolucent area is consistent with the presence of a placenta increta. Pasto and associates (1983) confirmed that the absence of a subplacental sonolucent or "hypoechoic retroplacental zone" is consistent with placenta increta.

More recently, magnetic resonance imaging has been used to diagnose placenta accreta (Maldjian and co-workers, 1999).

MANAGEMENT. The problems associated with delivery of the placenta and subsequent developments vary appreciably, depending upon the site of implantation, depth of myometrial penetration, and number of cotyledons involved. It is likely that focal placenta accreta with implantation in the upper uterine segment develops much more often than is recognized. The involved cotyledon is either pulled off the myometrium with perhaps somewhat excessive bleeding, or the cotyledon is torn from the placenta and adheres to the implantation site with increased bleeding, immediately or later. According to Benirschke and Kaufmann (2000), this may be one mechanism for formation of so-called placental polyps (Chap. 17, p. 406).

With more extensive involvement, hemorrhage becomes profuse as delivery of the placenta is attempted. In some cases, the placenta invades the broad ligament and entire cervix (Lin and co-workers, 1998). Such a case is shown in Figure 25-19. Successful treatment depends upon immediate blood replacement therapy as described on page 652 and nearly always prompt hysterectomy. Alternative measures are discussed subsequently (p. 647), and include uterine or internal iliac artery ligation or angiographic embolization. Scarantino and associates (1999) described use of argon beam coagulation for hemostasis in the lower uterine segment.

With total placenta accreta, there may be very little or no bleeding, at least until manual placental removal is attempted. At times, traction on the umbilical cord will invert the uterus, as described in the next section. Moreover, usual attempts at manual removal will not succeed because a cleavage plane between the maternal placental surface and the uterine wall cannot be developed. In the past, the most common form of "conservative" management was manual removal of as much placenta as possible and then packing of the uterus. In the review by Fox (1972), 25 percent of women managed conservatively died. Thus, the safest treatment in this circumstance is prompt hysterectomy.

INVERSION OF THE UTERUS

Complete uterine inversion after delivery of the infant is almost always the consequence of strong traction on an umbilical cord attached to a placenta implanted in the fundus (Fig. 25-21). Incomplete uterine inversion may also occur (Fig. 25-22). Contributing to uterine inversion is a tough cord that does not readily break away from the placenta, combined with fundal pressure and a relaxed uterus, including the lower segment and cervix. Placenta accreta may be implicated, although uterine inversion can occur without the placenta being so firmly adherent.

Shah-Hosseini and Evrard (1989) reported an incidence of about 1 in 6400 deliveries at the Women and Infants Hospital of Rhode Island. Of the 11 inversions identified, most were in primiparous women; immediate vaginal replacement of the inverted uterus was successful in nine instances. Platt and Druzin (1981) reported 28 cases in over 60,000 deliveries, for an incidence of about 1 in 2100. On the busy obstetrical service at Parkland Hospital, we encounter several cases annually and the vast majority are in "low-risk" deliveries.

CLINICAL COURSE. Uterine inversion is most often associated with immediate life-threatening hemorrhage, and without prompt treatment it may be fatal (Fig. 25-23). In the past it was stated that shock tends to be disproportionate to blood loss. Careful evaluation of the effects from transfusion of large volumes of blood in such cases does not support this concept, but instead makes it very apparent that blood loss in such circumstances was often massive but greatly underestimated (Watson and associates, 1980).

TREATMENT. Delay in treatment increases the mortality rate appreciably. It is imperative that a number of steps be taken immediately and simultaneously:

1. Assistance, including an anesthesiologist, is summoned immediately.

2. The freshly inverted uterus with placenta already separated from it may often be replaced simply by immediately pushing up on the fundus with the palm of the hand and fingers in the direction of the long axis of the vagina.

3. Preferably two intravenous infusion systems are made operational, and lactated Ringer solution and blood are given to treat hypovolemia.

4. If attached, the placenta is not removed until the infusion systems are operational, fluids are being given, and anesthesia, preferably halothane or enflurane, has been administered. Tocolytic drugs such as terbutaline, ritodrine, or magnesium sulfate have been used successfully for uterine relaxation and repositioning (Catanzarite and associates, 1986; Kovacs and DeVore, 1984; Thiery and Delbeke, 1985). In the meantime, the inverted uterus, if prolapsed beyond the vagina, is replaced within the vagina.

5. After removing the placenta, the palm of the hand is placed on the center of the fundus with the fingers extended to identify the margins of the cervix. Pressure is then applied with the hand so as to push the fundus upward through the cervix.

6. As soon as the uterus is restored to its normal configuration, the agent used to provide relaxation is stopped and simultaneously oxytocin is started to contract the uterus while the operator maintains the fundus in normal relationship.

Initially, bimanual compression (Fig. 25-18) will aid in control of further hemorrhage until uterine tone is recovered. After the uterus is well contracted, the operator continues to monitor the uterus transvaginally for any evidence of subsequent inversion.

SURGICAL INTERVENTION. Most often, the inverted uterus can be restored to its normal position by the techniques described. If the uterus cannot be reinverted by vaginal manipulation because of a dense constriction ring (Fig. 25-24), laparotomy is imperative. The fundus then may be simultaneously pushed upward from below and pulled from above. A traction suture well placed in the inverted fundus may be of aid. If the constriction ring still prohibits reposition, it is carefully incised posteriorly to expose the fundus. This surgical technique was described by Van Vugt and associates (1981). After replacement of the fundus, the anesthetic agent used to relax the myometrium is stopped, oxytocin infusion is begun, and the uterine incision repaired.

GENITAL TRACT LACERATIONS

PERINEAL LACERATIONS. All except the most superficial perineal lacerations are accompanied by varying degrees of injury to the lower portion of the vagina. Such tears may reach sufficient depth to involve the anal sphincter and may extend to varying depths through the walls of the vagina. Bilateral lacerations into the vagina are usually unequal in length and separated by a tongue-shaped portion of vaginal mucosa (see Fig. 13-17). Their repair should form part of every operation for the restoration of a lacerated perineum. If the underlying perineal and vaginal fascia and muscle are not sutured, this may lead to relaxation of the vaginal outlet and may contribute to rectocele and cystocele formation.

VAGINAL LACERATIONS. Isolated lacerations involving the middle or upper third of the vagina but unassociated with lacerations of the perineum or cervix are observed less commonly. These are usually longitudinal and frequently result from injuries sustained during a forceps or vacuum operation, but they may even develop with spontaneous delivery. Such lacerations frequently extend deep into the underlying tissues and may give rise to significant hemorrhage, which usually is controlled by appropriate suturing. They may be overlooked unless thorough inspection of the upper vagina is performed. Bleeding while the uterus is firmly contracted is strong evidence of genital tract laceration, retained placental fragments, or both.

Lacerations of the anterior vaginal wall in close proximity to the urethra are relatively common. They are often superficial with little to no bleeding, and repair is usually not indicated. If such lacerations are large enough to require extensive repair, difficulty in voiding can be anticipated and an indwelling catheter placed.

INJURIES TO LEVATOR ANI. These are the result of overdistention of the birth canal. Muscle fibers are separated and diminution in their tonicity may be sufficient to interfere with the function of the pelvic diaphragm. In such cases, the woman may develop pelvic relaxation. If these injuries involve the pubococcygeus muscle, urinary incontinence also may develop (Chap. 13, p. 325).

INJURIES TO THE CERVIX. The cervix is lacerated in over half of vaginal deliveries (Fahmy and associates, 1991). Most of these are less than 0.5 cm. Deep cervical tears may extend to the upper third of the vagina. In rare instances, however, the cervix may be entirely or partially avulsed from the vagina, with colporrhexis in the anterior, posterior, or lateral fornices. Such injuries sometimes follow difficult forceps rotations or deliveries performed through an incompletely dilated cervix with the forceps blades applied over the cervix. Rarely, cervical tears may extend to involve the lower uterine segment and uterine artery and its major branches, and even through the peritoneum. They may be totally unsuspected, but much more often they become manifest by excessive external hemorrhage or by hematoma formation. Extensive tears of the vaginal vault should be explored carefully. If there is question of peritoneal perforation, or of retroperitoneal or intraperitoneal hemorrhage, laparotomy should be considered. With damage of this severity, intrauterine exploration for possible rupture is also mandatory. Surgical repair is usually required, and effective anesthesia, vigorous blood replacement, and capable assistance are mandatory.

Cervical lacerations up to 2 cm must be regarded as inevitable in childbirth. Such tears heal rapidly and are rarely the source of any difficulty. In healing, they cause a significant change in the round shape of the external os before cervical effacement and dilatation to that of appreciable lateral elongation after delivery. As the consequence of such tears, there may be eversion of the cervix with exposure of the delicate mucus-producing endocervical glands.

Occasionally, during labor the edematous anterior lip of the cervix may be caught and compressed between the head and the symphysis pubis. If ischemia is severe, the cervical lip may undergo necrosis and separation. More rarely, the entire vaginal portion may be avulsed from the rest of the cervix. Such annular or circular detachment of the cervix is uncommon in modern obstetrics.

DIAGNOSIS. A deep cervical tear should always be suspected in cases of profuse hemorrhage during and after third-stage labor, particularly if the uterus is firmly contracted. Thorough examination is necessary, and the flabby cervix often makes digital examination alone unsatisfactory. Thus, the extent of the injury can be fully appreciated only after adequate exposure and visual inspection of the cervix. The best exposure is gained by the use of right-angle vaginal retractors by an assistant while the operator grasps the patulous cervix with a ring forceps (Fig. 25-25).

In view of the frequency with which deep tears follow major operative procedures, the cervix should be inspected routinely at the conclusion of the third stage after all difficult deliveries, even if there is no bleeding.

TREATMENT. Deep cervical tears require surgical repair. When the laceration is limited to the cervix, or even when it extends somewhat into the vaginal fornix, satisfactory results are obtained by suturing the cervix after bringing it into view at the vulva. Visualization is best accomplished when an assistant makes firm downward pressure on the uterus while the operator exerts traction on the lips of the cervix with fenestrated ovum or sponge forceps. Right-angle vaginal wall retractors are often helpful (Fig. 25-25). Because the hemorrhage usually comes from the upper angle of the wound, the first suture is applied just above the angle and sutured outward toward the operator. Associated vaginal lacerations may be tamponaded with gauze packs to retard hemorrhage while cervical lacerations are repaired. Either interrupted or running absorbable sutures are suitable. Overzealous suturing in an attempt to restore the normal cervical appearance may lead to subsequent stenosis during uterine involution.

PUERPERAL HEMATOMAS

From a review of six series, the incidence of puerperal hematomas was found to vary from 1 in 300 to 1 in 1000 deliveries (Gilstrap and colleagues, 2001). Nulliparity, episiotomy, and forceps delivery are the most commonly associated risk factors (Propst and Thorp, 1998; Ridgway, 1995). In many other cases, however, hematomas develop following injury to a blood vessel without laceration of the superficial tissues. These may develop with spontaneous or operative delivery. Occasionally, the hemorrhage is delayed.

Puerperal hematomas may be classified as vulvar, vulvovaginal, paravaginal, or retroperitoneal. Vulvar hematomas most often involve branches of the pudendal artery, including the posterior rectal, transverse perineal, or posterior labial artery, whereas paravaginal hematomas may involve the descending branch of the uterine artery (Zahn and Yeomans, 1990). Infrequently, the torn vessel lies above the pelvic fascia. In that event, the hematoma develops above it. In its early stages, the hematoma forms a rounded swelling that projects into the upper portion of the vaginal canal and may almost occlude its lumen. If the bleeding continues, it dissects retroperitoneally, and thus may form a tumor palpable above the Poupart ligament, or it may dissect upward, eventually reaching the lower margin of the diaphragm. Branches of the uterine artery may be involved with these types of hematomas.

VULVAR HEMATOMAS. These hematomas, such as the one shown in Figure 25-26, particularly those that develop rapidly, may cause excruciating pain, which often is the first symptom noticed. Hematomas of moderate size may be absorbed spontaneously. The tissues overlying the hematoma may give way as a result of necrosis caused by pressure, and profuse hemorrhage may follow. In other cases, the contents of the hematoma may be discharged in the form of large clots. In the subperitoneal variety, extravasation of blood beneath the peritoneum may be massive and occasionally fatal.

DIAGNOSIS. A vulvar hematoma is readily diagnosed by severe perineal pain and the sudden appearance of a tense, fluctuant, and sensitive tumor of varying size covered by discolored skin. When the mass develops adjacent to the vagina, it may escape detection temporarily; but symptoms of pressure, if not pain, or inability to void should prompt a vaginal examination with discovery of a round, fluctuant tumor encroaching on the lumen. When the hematoma extends upward between the folds of the broad ligament, it may escape detection unless a portion of the tumor can be felt on abdominal palpation or unless hypovolemia develops. These are worrisome because large hematomas have led to death.

TREATMENT. Smaller vulvar hematomas identified after leaving the delivery room may be treated expectantly (Propst and Thorp, 1998). If, however, the pain is severe, or if the hematoma continues to enlarge, the best treatment is prompt incision. This is done at the point of maximal distention along with evacuation of blood and clots and ligation of bleeding points. The cavity may then be obliterated with mattress sutures. Often, no sites of bleeding are identified after the hematoma has been drained. In such cases, the vagina—and not the hematoma cavity—is packed for 12 to 24 hours. With hematomas of the genital tract, blood loss is nearly always considerably more than the clinical estimate. Hypovolemia and severe anemia should be prevented by adequate blood replacement. In about half of women with hematomas requiring surgical repair, transfusions are necessary (Zahn and Yeomans, 1990).

Subperitoneal and supravaginal hematomas are more difficult to treat. They can be evacuated by incision of the perineum; but unless there is complete hemostasis, which is difficult to achieve by this route, laparotomy is advisable.

ANGIOGRAPHIC EMBOLIZATION. This technique has become popular for management of intractable puerperal hematomas. In can be used primarily, or usually when hemostasis is not obtained by surgical methods. A case in which embolization was carried out with occlusion of the internal pudendal artery and its vaginal branch as well as the uterine artery is shown in Figure 25-27. Alvarez and co-workers (1992) and Hsu and Wan (1998) have reviewed indications for angiographic embolization and described cases in which it was employed.

RUPTURE OF THE UTERUS

The incidence of uterine rupture may vary appreciably among institutions. According to the Centers for Disease Control and Prevention (2000), use of hospital discharge data is inaccurate for surveillance purposes. Although the frequency of uterine rupture from all causes has probably not decreased remarkably during the past several decades, the etiology of rupture has changed appreciably and the outcome has improved significantly. Still, maternal mortality is reported (Ripley, 1999). Indeed, 20 percent of maternal deaths from hemorrhage were due to ruptured uterus (Nagaya and colleagues, 2000).

Eden and associates (1986) reviewed experiences with uterine rupture over a 53-year period at Duke University. From 1931 to 1950 the incidence was 1 in 1280 deliveries compared with 1 in 2250 from 1973 to 1983. Rachagan and colleagues (1991) reported a similar incidence of about 1 in 3000 over a 21-year period. Miller and Paul (1996) reported an incidence of about 1 in 1235 in nearly 190,000 deliveries from Los Angeles County-University of Southern California Women's Hospital. More than 90 percent of these were associated with a prior cesarean delivery. In our experiences, true uterine rupture, as will be defined, has become exceedingly rare. For the 5-year period from 1990 through 1994, during which time nearly 74,000 women were delivered at Parkland Hospital, there were only four uterine ruptures for an incidence of 1 in 18,500 deliveries. This low rate is related to our policy of not inducing or augmenting labor with oxytocin in women with a prior cesarean delivery (Chap. 23, p. 540).

ETIOLOGY. Uterine rupture may develop as a result of preexisting injury or anomaly, it may be associated with trauma, or it may complicate labor in a previously unscarred uterus. A classification of the etiology of uterine rupture is presented in Table 25-7.

The most common cause of uterine rupture is separation of a previous cesarean section scar. As discussed in Chapter 23 (p. 542), this is increasing with the developing trend of allowing a trial of labor following prior transverse cesarean section(s). Farmer and colleagues (1991) reported that two thirds of over 11,000 women with a prior cesarean delivery underwent a trial of labor with an incidence of overt uterine rupture of 0.8 percent. In the study cited above by Miller and Paul (1996), only 11 of 153 cases of uterine rupture were not associated with prior cesarean section. Midtrimester uterine rupture is rare (Levrant and Wingate, 1996). However, women undergoing midpregnancy terminations following a prior cesarean were found to have a 3.8 percent risk of rupture (Chapman and associates, 1996).

Other common predisposing factors to uterine rupture are previous traumatizing operations or manipulations such as curettage, perforation, or myomectomy (Fedorkow and colleagues, 1987; Pelosi and Pelosi, 1997). Excessive or inappropriate uterine stimulation with oxytocin, a previously common cause, has become very uncommon. Mishra and colleagues (1995) described a 43-year-old woman who suffered a ruptured vertical cesarean incision associated with inhaled crack cocaine.

DEFINITIONS. Rupture of the uterus may communicate directly with the peritoneal cavity (complete) or may be separated from it by the visceral peritoneum over the uterus or that of the broad ligament (incomplete).

It is important to differentiate between rupture versus dehiscence of a cesarean section scar. Rupture refers to separation of the old uterine incision throughout most of its length, with rupture of the fetal membranes so that the uterine cavity and the peritoneal cavity communicate. In these circumstances, all or part of the fetus is usually extruded into the peritoneal cavity. In addition, there is usually significant bleeding from the edges of the scar or from an extension of the rent into previously uninvolved uterus. By contrast, with dehiscence of a cesarean section scar, the fetal membranes are not ruptured and the fetus is not extruded into the peritoneal cavity. Typically, with dehiscence, the separation does not involve all of the previous uterine scar, the peritoneum overlying the defect is intact, and bleeding is absent or minimal.

CLASSICAL VERSUS LOWER-SEGMENT CESAREAN SECTION SCARS. With prior cesarean delivery, the American College of Obstetricians and Gynecologists (1999) cite the following figures for a trial of labor and uterine rupture: 1 to 7 percent with a prior low-vertical cesarean incision, 4 to 9 percent with a T-shaped incision, and 4 to 9 percent with a classical scar. Importantly, in about a third of cases the classical scar ruptures before labor, and not infrequently, it takes place several weeks before term. We recently encountered a term abdominal pregnancy in a woman whose prior classical incision had separated weeks to months before she was delivered by repeat cesarean.

RUPTURE OF A CESAREAN SECTION SCAR. As discussed in detail in Chapter 23 (p. 540), the current trend is to offer and even encourage a trial of labor for women who had one previous transverse cesarean delivery (American College of Obstetricians and Gynecologists, 1999). The main drawback is that separation of the previous scar complicates about 1 in 200 trials of labor. This is higher in women with two or more prior operations who have a trial of labor, either spontaneous or with oxytocin stimulation.

The experience at Parkland Hospital has been that separation of the transverse uterine incision that develops antepartum or during early labor usually is limited to dehiscence without an appreciable increase in maternal or perinatal morbidity. From 1986 through 1990, there were a total of 7049 women at Parkland Hospital with prior cesarean sections, and 2044, or almost 30 percent, were allowed a trial of labor. Of the women undergoing such a trial, 1482 (73 percent) delivered vaginally. Uterine rupture with part of the fetus extruded outside of the uterus developed in three women, for a rate of 1.5 per 1000. In the entire group, there were two stillbirths (1 per 1000) and four women required hysterectomy (2 per 1000). In another 307 women undergoing a trial of labor who were given oxytocin, there were three uterine ruptures (10 per 1000). Since 1990, when we discontinued the use of oxytocin for labor induction or augmentation for women with prior cesarean deliveries, there have been no uterine ruptures in these women.

Separation of a vertical (classical) scar is more likely to result in severe hemorrhage with increased perinatal morbidity and mortality. It also more likely will require hysterectomy (Fig. 25-28).

MORBIDITY AND MORTALITY. Maternal risks are related to whether there is rupture of an intact uterus versus a prior cesarean scar. Scar separation following a trial of labor in a woman with a prior transverse incision has not been associated with maternal deaths (Flamm and colleagues, 1988; Rachagan and associates, 1991). Conversely, in the 24 cases of uterine rupture principally unassociated with prior incisions, Eden and associates (1986) reported one maternal death and a 46 percent perinatal loss. Likewise, Rachagan and colleagues (1991) reported fetal mortality to be almost 70 percent with either spontaneous or traumatic uterine rupture. As discussed in Chapter 23 (p. 542), perinatal morbidity and mortality can be substantive with rupture during labor of a prior uterine incision.

HEALING OF THE CESAREAN SECTION SCAR. There is little information that deals with healing of cesarean section scars. Williams (1921) believed that the uterus heals by regeneration of the muscular fibers and not by scar tissue formation. For example, upon inspection of the unopened uterus at repeat cesarean delivery, there is usually no trace of the former incision, or at most, an almost invisible linear scar. Also, when the uterus is removed and fixed in formalin, often no scar is visible, or only a shallow vertical furrow in the external and internal surfaces of the anterior uterine wall is seen, with no trace of scar tissue between them.

Schwarz and co-workers (1938) concluded that healing occurs mainly by fibroblast proliferation. They studied the uterine incision site some days after cesarean incision and observed that as the scar shrinks, connective tissue proliferation becomes less obvious. If the cut surfaces of the uterus are closely apposed, the proliferation of connective tissue is minimal, and the normal relation of smooth muscle to connective tissue is gradually reestablished. Even when the healing is so poor that marked thinning has resulted, the remaining tissue often is entirely muscular.

RUPTURE OF THE UNSCARRED UTERUS

TRAUMATIC RUPTURE. Although the uterus is surprisingly resistant to blunt trauma, pregnant women sustaining blunt trauma to the abdomen should be watched carefully for signs of a ruptured uterus (Chap. 43, p. 1173). Miller and Paul (1996) described only three cases due to trauma in over 150 women with a ruptured uterus. The likelihood of placental abruption with blunt trauma is greater. Conversely, penetrating abdominal wounds are much more likely to involve the large pregnant uterus.

In the past, traumatic rupture during delivery often was caused by internal podalic version and extraction. Other causes of traumatic rupture include difficult forceps delivery, breech extraction (Fig. 25-29), and unusual fetal enlargement, such as hydrocephaly.

SPONTANEOUS RUPTURE. In the study cited earlier by Miller and Paul (1996), the incidence of spontaneous uterine rupture was only about 1 in 15,000 deliveries. They also found that rupture is more likely in women of high parity (Miller and colleagues, 1997). Oxytocin stimulation of labor has been rather commonly associated with uterine rupture, especially in women of high parity (Fuchs and co-workers, 1985; Rachagan and associates, 1991). Maymon and associates (1991) as well as Bennett (1997) have reported uterine rupture with labor induction using prostaglandin E₂ gel or E₁ vaginal tablets. For these reasons, oxytocin should be given with great caution to stimulate labor in women of high parity. Similarly, in women of high parity, a trial of labor in the presence of suspected cephalopelvic disproportion, or abnormal presentation such as a brow, must be undertaken with caution.

PATHOLOGICAL ANATOMY. The role in uterine rupture of excessive stretching of the lower uterine segment with the development of a pathological retraction ring is described in Chapter 18. Rupture of the previously intact uterus at the time of labor most often involves the thinned-out lower uterine segment. The rent, when it is in the immediate vicinity of the cervix, frequently extends transversely or obliquely. Usually the tear is longitudinal when it occurs in the portion of the uterus adjacent to the broad ligament (Fig. 25-29). Although developing primarily in the lower uterine segment, it is not unusual for the laceration to extend further upward into the body of the uterus or downward through the cervix into the vagina. At times, the bladder may also be lacerated (Rachagan and colleagues, 1991). After complete rupture, the uterine contents escape into the peritoneal cavity, unless the presenting part is firmly engaged, when only a portion of the fetus may be extruded from the uterus.

With uterine rupture in which the peritoneum remains intact, hemorrhage frequently extends into the broad ligament. In such circumstances, hemorrhage tends to be less severe than with intraperitoneal rupture. Such bleeding results in a large retroperitoneal hematoma that may involve sufficient blood loss to cause death. Fatal exsanguination may also supervene after rupture of the hematoma relieves the tamponading effect of the intact broad ligament.

CLINICAL COURSE. Prior to circulatory collapse from hemorrhage, the symptoms and physical findings may appear bizarre unless the possibility of uterine rupture is kept in mind. For example, hemoperitoneum from a ruptured uterus may result in irritation of the diaphragm and pain referred to the chest—leading one to a diagnosis of pulmonary or amnionic fluid embolus instead of uterine rupture.

Although once taught, it appears that few women experience cessation of contractions following uterine rupture. Rodriguez and colleagues (1989) described data from 39 of 76 women with uterine rupture in whom there was an intrauterine pressure catheter. In none of these was there a loss of intrauterine pressure or cessation of labor. Four women had increased baseline pressure associated with severe variable decelerations (Fig. 25-30). The most common finding in all of these 76 cases was sudden, severe fetal heart rate decelerations, seen in almost 80 percent of cases. They concluded that intrauterine pressure monitoring added little to the diagnosis of uterine rupture.

In other women, the appearance is identical to that of placental abruption. In still others, rupture is unaccompanied by appreciable pain and tenderness. Also, because most women in labor are given something for discomfort (either narcotics or lumbar epidural analgesia), pain and tenderness may not be readily apparent and the condition becomes evident because of signs of fetal distress, maternal hypovolemia from concealed hemorrhage, or both.

In some cases in which the fetal presenting part had entered the pelvis with labor, there is loss of station detected by pelvic examination. If the fetus is partly or totally extruded, abdominal palpation or vaginal examination is helpful to identify the presenting part, which has moved away from the pelvic inlet. A firm contracted uterus may at times be felt alongside the fetus. Often fetal parts are more easily palpated than usual. On vaginal examination, it is sometimes possible to palpate a tear in the uterine wall through which the fingers can be passed into the peritoneal cavity. Failure to detect the tear by no means proves its absence. In suspected cases, it is imperative that thorough examination be performed by an experienced examiner before the suspicion is abandoned. Sonography, performed on site, may be useful.

PROGNOSIS. With rupture and expulsion of the fetus into the peritoneal cavity, the chances for intact fetal survival are dismal, and mortality rates reported in various studies range from 50 to 75 percent. If the fetus is alive at the time of the rupture, the only chance of continued survival is afforded by immediate delivery, most often by laparotomy. Otherwise, hypoxia from both placental separation and maternal hypovolemia is inevitable. If untreated, most of the women die from hemorrhage or, less often, later from infection. Prompt diagnosis, immediate operation, the availability of large amounts of blood, and antimicrobial therapy have greatly improved the prognosis. Conversely, Nkata (1996) reported 44 percent maternal deaths in 32 women with a ruptured uterus in Zambia.

HYSTERECTOMY VERSUS REPAIR. In cases of scar separation without bleeding following vaginal birth after cesarean section, exploratory laparotomy is not indicated. With spontaneous uterine rupture, or frank rupture during a trial of labor after cesarean delivery, hysterectomy is frequently required. In two recent reports by McMahon (1996) and Miller (1997) and their colleagues, 10 to 20 percent of these women required hysterectomy for hemostasis. In selected cases, suture of the wound with uterine preservation may be performed. Sheth (1968) described outcomes from a series of 66 cases in which repair of a uterine rupture was elected rather than hysterectomy. In 25 instances, the repair was accompanied by tubal sterilization. Thirteen of the 41 mothers who did not have tubal sterilization had a total of 21 subsequent pregnancies, but uterine rupture recurred in four instances. Martin and colleagues (1990) described a woman in whom recurrent uterine lateral fundal separation at 19 weeks was repaired using a Gore-Tex soft tissue patch. When delivered by cesarean section at 33 weeks, the patch was intact and epithelialized.

In the presence of a large hematoma in the broad ligament, identification and ligation of the uterine vessels can be extremely difficult. In general, efforts to control hemorrhage by clamping indiscriminately at the site of rupture involving the lower segment should be avoided. To do otherwise often leads to clamping and ligation of the ureter, bladder, or both. With uterine ruptures involving the lower uterine segment, bleeding vessels must be visualized free of surrounding tissue before clamping, or the ureter and bladder must be demonstrated to be remote from the tissue that is clamped. In some cases, the transected uterine artery has retracted laterally and is displaced to the pelvic sidewall by the hematoma that resulted. Placement of clamps to control bleeding carries little risk when rupture involves the body of the uterus remote from the ureters and bladder. The broad ligament may be entered and the ascending uterine artery and veins safely clamped. Usually, the ovarian vessels should be promptly clamped adjacent to the uterus.

INTERNAL ILIAC ARTERY LIGATION. Ligation of the internal iliac arteries at times reduces the hemorrhage appreciably (Allahbadia, 1993; Clark and colleagues, 1985). This operation is more easily performed if the midline abdominal incision is extended upward above the umbilicus. With adequate exposure, ligation is accomplished by opening the peritoneum over the common iliac artery and dissecting down to the bifurcation of the external and internal iliac arteries. The areolar sheath covering the internal iliac artery is incised longitudinally, and a right-angle clamp is carefully passed just beneath the artery. Care must be taken not to perforate contiguous large veins, especially the internal iliac vein. Suture, usually nonabsorbable, is then inserted into the open clamp, the jaws are locked, the suture is carried around the vessel, and the vessel is securely ligated (Figs. 25-31 and 25-32). Pulsations in the external iliac artery, if present before tying the ligature, should be present afterward as well. If not, pulsations must be identified after arterial hypotension has been successfully treated, in order to assure that the blood flow through the external iliac vessel has not been compromised by the ligature. The most important mechanism of action with internal iliac artery ligation is an 85 percent reduction in pulse pressure in those arteries distal to the ligation (Burchell, 1968), thus turning an arterial pressure system into one with pressures approaching those in the venous circulation and more amenable to hemostasis via simple clot formation. Bilateral ligation of these arteries does not appear to interfere seriously with subsequent reproduction. Mengert and associates (1969) documented successful pregnancies in five women after bilateral internal iliac artery ligation. In three, the ovarian arteries were also ligated.

In some women, pelvic vessel bleeding may continue even after internal iliac artery ligation. Angiographically directed arterial embolization has been reported to successfully arrest hemorrhage. This was described on page 647 and in Figure 25-27.

HYPOVOLEMIC SHOCK

Shock from hemorrhage evolves through several stages. Early in the course of massive bleeding, there are decreases in mean arterial pressure, stroke volume, cardiac output, central venous pressure, and pulmonary capillary wedge pressure. Increases in arteriovenous oxygen content difference reflect a relative increase in tissue oxygen extraction, although overall oxygen consumption falls (Bland and colleagues, 1985).

Blood flow to capillary beds in various organs is controlled by arterioles, which are resistance vessels that in turn are controlled by the central nervous system. At least 70 percent of total blood volume is contained in venules, which are passive resistance vessels controlled by humoral factors. Catecholamine release during hemorrhage causes a generalized increase in venular tone, resulting in an autotransfusion from this capacitance reservoir (Barber and colleagues, 1999). These changes are accompanied by compensatory increases in heart rate, systemic and pulmonary vascular resistance, and myocardial contractility. In addition, there is redistribution of cardiac output and blood volume by selective centrally mediated arteriolar constriction. This results in diminished perfusion to the kidneys, splanchnic beds, skin, and uterus with relative maintenance of blood flow to the heart, brain, and adrenal glands, organs that auto-regulate their own flow (Barber and associates, 1999).

As blood volume deficit exceeds 25 percent, compensatory mechanisms usually are inadequate to maintain cardiac output and blood pressure. At this point, additional small losses of blood result in rapid clinical deterioration. Despite an initial increase in total oxygen extraction by maternal tissue, maldistribution of blood flow results in local tissue hypoxia and metabolic acidosis, producing a vicious cycle of vasoconstriction, organ is-chemia, and cellular death. Hemorrhage also activates the CD-18 locus of lymphocytes and monocytes, which mediates leukocyte-endothelial cell interactions. These events lead to loss of capillary membrane integrity and additional loss of intravascular volume. A number of these adverse effects appear to be mediated by peptide leukotrienes and cytokines and may be experimentally improved by their pharmacological antagonism mediators (Bitterman and colleagues, 1988). There is also increased platelet aggregation in hypovolemic shock, resulting in the release of a number of vasoactive mediators that cause small vessel occlusion and further impairment of microcirculatory perfusion.

Often overlooked is the importance of extracellular fluid and electrolyte shifts in both pathophysiology and successful treatment of hypovolemic shock. This involves changes in the cellular transport of various ions, in which sodium and water enter skeletal muscles and cellular potassium is lost to the extracellular fluid (Chiao and colleagues, 1990). Replacement of extracellular fluid is thus an important component of therapy in hypovolemic shock. Indeed, survival appears to be reduced in acute hemorrhagic shock when blood alone—compared with blood and lactated Ringer solution—is administered (Barber and co-workers, 1999).

ESTIMATION OF BLOOD LOSS

Visual inspection is most often used but is notoriously inaccurate. In some reports, the amount of blood estimated to have been lost by inspection was on average about half the measured loss. Importantly, in obstetrics, part or all of the hemorrhage may be concealed. It is important to realize that in a situation of acute hemorrhage, the immediate hematocrit may not reflect actual blood loss. After the loss of 1000 mL, the hematocrit typically falls only 3 volumes percent in the first hour. However, rapid infusion of intravenous crystalloids will allow for more rapid equilibration.

Urine output is one of the most important parameters to follow in the bleeding patient. When carefully measured, the rate of urine formation, in the absence of diuretics, reflects the adequacy of renal perfusion and, in turn, perfusion of other vital organs, because renal blood flow is especially sensitive to blood volume changes. Urine flow of at least 30 mL and preferably 60 mL per hour should be maintained. With potentially serious hemorrhage, an indwelling catheter should be inserted promptly to measure urine flow. Potent diuretics such as furosemide invalidate the relationship between urine flow and renal perfusion. This need not be a problem in the management of the woman who is hemorrhaging, however, because the use of diuretics is contraindicated. Further intravascular volume reduction with diuretics is harmful in a hypovolemic patient. Another effect of furosemide is venodilation, which further reduces cardiac venous return, thereby further compromising cardiac output.

RESUSCITATION AND ACUTE MANAGEMENT

Whenever there is any suggestion of excessive blood loss, it is essential that steps be immediately taken to identify the presence of uterine atony, retained placental fragments, or genital tract lacerations. It is imperative that at least one or two intravenous infusion systems of large caliber be established immediately to allow rapid administration of crystalloid solutions and blood. An operating room, surgical team, and anesthesiologist should always be immediately available. The management of specific causes of postpartum hemorrhage were discussed earlier in this chapter.

FLUID AND BLOOD REPLACEMENT. Treatment of serious hemorrhage demands prompt and adequate refilling of the intravascular compartment. Crystalloid solutions typically are used for initial volume resuscitation. Such solutions rapidly equilibrate into the extravascular space and only 20 percent of crystalloid remains in the circulation of critically ill patients after 1 hour (Shoemaker and co-workers, 1991). Because of such equilibration, initial fluid infusion should involve about three times as much crystalloid as the estimated blood loss.

There is debate concerning fluid resuscitation of hypovolemic shock with colloid versus crystalloid solutions. According to their review, Schierhout and Roberts (1998) found a 4-percent excessive mortality in nonpregnant patients resuscitated with colloid compared with crystalloid. The Cochrane Injuries Group Albumin Reviewers (1998) found a 6-percent excess mortality in albumin-treated nonpregnant patients with shock. We concur with Bonnar (2000) that fluid resuscitation preferably should be with crystalloid and blood.

Considerable debate also surrounds the hematocrit level or hemoglobin concentration that mandates blood transfusion. According to deliberations of a Consensus Development Conference (1988), cardiac output does not substantively increase until the hemoglobin concentration falls to about 7 g/dL. Although the committee reported that otherwise healthy anesthetized animals survived isovolemic anemia with hematocrit decreases down to 5 volume percent, they further cited that there was significant functional deterioration well before that point. It is difficult to define a universal hematocrit or hemoglobin value below or above which transfusion is either mandatory or contraindicated. However, the recommendations of the Consensus Development Conference should be considered in clinical decision making. According to these guidelines, red blood cells are not infused for moderate anemia in stable women.

For the woman acutely bleeding, we recommend rapid blood infusion if the hematocrit is less than 25 volume percent. Similarly, Morrison and colleagues (1993) recommend transfusion if the hematocrit is less than 24 volume percent or if hemoglobin is less than 8 g/dL if there is imminent surgery, acute operative blood loss, acute hypoxia, vascular collapse, or other factors present. Further support for these recommendations was provided by Czer and Shoemaker (1978). In 94 critically ill postoperative patients, mortality rates were lowest when hematocrit values were maintained between 27 and 33 volume percent.

Hebert and associates (1999) reported results from the Canadian Critical Care Trials Group. A total of 838 critically ill nonpregnant patients were randomized to restrictive red cell transfusions to maintain hemoglobin concentration over 7 g/dL, or to liberal transfusions to maintain the hemoglobin 10 to 12 g/dL. The 30-day mortality rate was similar (19 versus 23 percent, restrictive versus liberal), however, in the less ill patients (APACHE score 20 or less) the 30-day mortality was significantly lower in the restrictive group (9 versus 26 percent). Morrison and colleagues (1991) reported no benefits of red cell transfusions given to women who had suffered postpartum hemorrhage and who were isovolemic but anemic with a hematocrit between 18 and 25 volume percent. Clearly, the level to which a woman is transfused depends not only on the present red cell mass, but also on the likelihood of additional blood loss.

BLOOD AND COMPONENT REPLACEMENT. Contents and effects of transfusion of various blood components are shown in Table 25-8. Compatible whole blood is ideal for treatment of hypovolemia from catastrophic acute hemorrhage. It has a shelf life of 40 days, and 70 percent of the transfused red cells remain viable for at least 24 hours following transfusion. It replaces many coagulation factors, and especially fibrinogen, and its plasma expands the hypovolemia from hemorrhage. Overall, the bleeding patient is resuscitated with fewer blood donor exposures with whole blood. One unit of whole blood will raise the hematocrit by 3 to 4 volume percent. For two decades, in most cases of obstetrical hemorrhage, as well as most every other field of medicine, red cell replacement has been used and usually proves to be sufficient. The exception is the woman with torrential bleeding.

Autotransfusion has been used for many years in surgical procedures and is becoming more popular (Schwartz, 1999). The safety of intraoperative autologous blood salvage and autotransfusion was evaluated in a multicenter historical cohort study by Rebarber and colleagues (1998). When 139 women undergoing cesarean delivery and given autotransfusion were compared with 87 control women, there were no differences in adverse outcomes. Specifically, there was no evidence for respiratory distress or amnionic fluid embolism.

Fractionation of whole blood makes available specific components—clotting factors and platelets—that otherwise would be scarce and thus unavailable for specific deficiencies. According to the National Institutes of Health (1993), component therapy provides better treatment because only the specific component needed is given. It also conserves blood resources because components from one unit of blood can be used for several patients. For these reasons, the infusion of whole banked blood is usually not necessary and rarely, available (Barber and colleagues, 1999; Bonnar, 2000; Klein, 1994).

DILUTIONAL COAGULOPATHY. When blood loss is massive, replacement with crystalloid solutions and packed red blood cells usually results in a depletion of platelets and soluble clotting factors, leading to a functional coagulopathy that clinically is indistinguishable from disseminated intravascular coagulopathy (p. 657). This impairs hemostasis and further contributes to blood loss. The most frequent coagulation defect found in women with blood loss and multiple transfusions is thrombocytopenia (Counts and associates, 1979; Wilson and associates, 1971). Because stored whole blood is deficient in factors V, VIII, XI, and platelets, and all soluble clotting factors are absent from packed red blood cells, severe hemorrhage without factor replacement may also cause hypofibrinogenemia and prolongation of the prothrombin and partial thromboplastin times. In some cases, frank consumptive coagulopathy may accompany shock and confuse the distinction between dilutional and consumptive coagulopathy. Fortunately, in most situations encountered in obstetrics, treatment of both is the same.

Although various algorithms have been proposed to guide the replacement of platelets and clotting factors according to the volume of blood loss, there is great patient variability. Thus, component replacement is rarely necessary with acute replacement of 5 to 10 units of packed red blood cells or less. However, when blood loss exceeds this amount, consideration should be given to laboratory evaluation of platelet count, clotting studies, and fibrinogen levels. Fortunately, in practice the level of various clotting factors required for adequate hemostasis is quite minimal.

In the bleeding woman, the platelet count should be maintained above 50,000/uL with the infusion of platelet concentrates. A fibrinogen level of less than 100 mg/dL or sufficiently prolonged prothrombin or partial thromboplastin times in a patient with surgical bleeding is an indication for fresh-frozen plasma administration in doses of 10 to 15 mL/kg. If fibrinogen is severely depleted and the woman is bleeding, 15 units of cryoprecipitate given rapidly will result in plasma levels above 100 mg/dL.

TYPE AND SCREEN VERSUS CROSS-MATCH. Blood transfusion is usually not needed in most women delivered vaginally, and even with cesarean section, only 2 to 5 percent will require transfusion (Dickason and Dinsmoor, 1992; Klapholtz, 1990). In any women at significant risk for hemorrhage, typing and screening or cross-matching is essential. The screening procedure involves mixing the maternal serum with standard reagent red cells that contain the antigens with which most of the common clinically significant antibodies will react. A cross-match, on the other hand, involves the use of actual donor erythrocytes rather than standard red cells. Only 0.03 to 0.07 percent of patients who are determined not to have antibodies in a type and screen procedure subsequently will have antibodies as determined by cross-match (Boral and colleagues, 1979). Thus, in an emergency, administration of screened blood very rarely results in adverse clinical sequela. Not testing for a cross-match also decreases blood bank costs. Further, blood that is cross-matched is held exclusively for a single potential recipient, whereas with screening, blood is available for more than one potential recipient and wastage of banked blood is reduced. For these reasons, type and screen is preferred in most obstetrical situations.

PACKED RED BLOOD CELLS. Cells packed from a unit of whole blood have a hematocrit of 60 to 70 volumes percent, depending upon the method used for preparation and storage. A unit of packed red blood cells contains the same volume of erythrocytes as whole blood, and will also raise the hematocrit by 3 to 4 volume percent. Packed red blood cell and crystalloid infusion are the mainstays of transfusion therapy for most cases of obstetrical hemorrhage.

PLATELETS. When transfusion is needed, it is preferable to give platelets obtained by apheresis from one donor. In this scheme, the equivalent of platelets from six individual donors is given as a one-unit one-donor transfusion. Such units generally cannot be stored more than 5 days. The donor plasma must be compatible with recipient erythrocytes. Further, because some red blood cells are invariably transfused along with the platelets, only platelets from D-negative donors should be given to D-negative recipients. Platelet transfusion is considered in a bleeding patient with a platelet count below 50,000/uL. In the nonsurgical patient, bleeding is rarely encountered if the platelet count exceeds 5000 to 10,000/uL (Sachs, 1991).

If single-donor platelets are not available, random donor platelet packs are used. These are prepared from individual units of whole blood by centrifugation, then resuspended in 50 to 70 mL of plasma. One unit of random donor platelets contains about 5.5 ´ 10¹⁰ platelets; 6 to 10 such units are generally transfused. Each unit transfused should raise the platelet count by 5000/uL (National Institutes of Health, 1993).

FRESH-FROZEN PLASMA. This component is prepared by separating plasma from whole blood and then freezing it. It is a source of all stable and labile clotting factors, including fibrinogen. It is often used in the acute treatment of women with consumptive or dilutional co-agulopathy. Fresh-frozen plasma is not appropriate for use as a volume expander in the absence of specific clotting factor deficiency. It should be considered in a bleeding woman with a fibrinogen level below 100 mg/dL and abnormal prothrombin and partial thromboplastin times.

CRYOPRECIPITATE. This component is prepared from fresh-frozen plasma. It contains factor VIII: C, factor VIII: von Willebrand factor, fibrinogen (at least 150 mg), factor XIII, and fibronectin in less than 15 mL of plasma from which it was derived (American Association of Blood Banks, 1994). There is no advantage to the use of cryoprecipitate for general clotting factor replacement in the bleeding woman instead of fresh-frozen plasma. Cryoprecipitate is only indicated in states of general factor deficiency where potential volume overload is a problem, and in a few conditions involving deficiency of specific factors. A major indication for this fraction is with severe hypofibrinogenemia caused by placental abruption in a woman with surgical incisions.

AUTOLOGOUS TRANSFUSIONS. Under some circumstances, autologous blood storage for transfusion may be considered. McVay and colleagues (1989) reported observations from 273 pregnant women in whom blood was drawn in the third trimester. Minimal requirements were a hemoglobin concentration 11 g/dL or a hematocrit of 34 volume percent. However, almost three fourths of these women donated only one unit, a volume of questionable value. They reported no complications. Further, the need for transfusion generally cannot be predicted. Sherman and colleagues (1992) studied 27 women given two or more transfusions in over 16,000 deliveries. In only 40 percent was an antepartum risk factor identified. Andres and co-workers (1990) and Etchason and associates (1995) concluded that autologous transfusions were not cost effective.

TRANSFUSION-RELATED INFECTIONS. With each unit of blood or any component thereof, the recipient is exposed to the risk of blood-borne infections. In a multihospital prospective follow-up done in London, Regan and colleagues (2000) tested almost 5600 recipients of nearly 22,000 units of blood. All were seronegative when tested 9 months after transfusion for hepatitis B and C, human immunodeficiency virus (HIV) infection, and human T-cell leukemia/lymphoma virus (HTLV) types I and II.

Fortunately, the most feared infection—human immunodeficiency virus (HIV)—is the least common. Earlier estimates with donor screening for viral antibody placed the risk of infection at 1 in 40,000 to 1 in 310,000 (National Institutes of Health, 1993). Unfortunately, more than 60 percent of recipients of HIV-positive blood become seropositive, and half develop acquired immunodeficiency syndrome (AIDS) within 7 years (Ward and associates, 1988). Because of improved sensitivity of enzyme immunosorbent assays, the risk of HIV transmission in screened blood is currently computed to be 1 in 500,000 to 1 million donations (Cohn, 2000; Lackritz and associates, 1995; Schreiber and co-workers, 1996).

The likelihood of HIV-2 infection is even less. After implementation of a combined HIV-1/HIV-2 screening of blood donors in 1992, only three units of 74 million tested through 1995 were positive for HIV-2 (Centers for Disease Control and Prevention, 1995).

Until relatively recently, the transmission of non-A, non-B hepatitis virus was much more likely to complicate transfusions. The prevalence of hepatitis C is 1 to 2 percent of donors. In the past, most cases were undetected because they caused anicteric infection, although chronic hepatitis was common (Chap. 48, p. 1289). Serological testing for hepatitis C antibody became available in 1990 and the American Association of Blood Banks mandated hepatitis C testing for all donors. With current testing techniques, the risk of hepatitis C transmission is approximately 1 in 3300 to 1 in 103,000 (Schreiber and associates, 1996; Sloand and colleagues, 1995). The risk of transmitting other infectious disease with transfusion, such as malaria and cytomegalovirus, is estimated to be less than 1 in 1 million (National Institutes of Health, 1993).

RED-CELL SUBSTITUTES. These are of three varieties: perfluorochemicals, liposome-encapsulated hemoglobin, and hemoglobin-based oxygen carriers. Their history and development was recently reviewed by Cohn (2000). Fluoridated hydrocarbons are biologically inert liquids with relatively high oxygen solubility. The use of such emulsions allows oxygen to be transported and delivered to tissues by simple diffusion. The most commonly used emulsion, Fluosol, must be stored frozen and thawed within 24 hours of use. Clinical benefits of these emulsions are not well established, but they may decrease the need for blood with extensive hemorrhage (Klein, 2000). Liposome-encapsulated hemoglobin has not proved promising. One formulation of a hemoglobin-based oxygen carrier, diaspirin cross-linked hemoglobin (DCLHb), proved to be dangerous (Sloan and colleagues, 1999). More recently, Mullon and co-workers (2000) have described successful use of polymerized bovine hemoglobin, HBOC-201, as a blood substitute for a nonpregnant women with severe autoimmune hemolytic anemia.

CONSUMPTIVE COAGULOPATHY

In 1901, DeLee reported that "temporary hemophilia" developed in a woman with a placental abruption and another with a long-dead macerated fetus. Observations that extensive placental abruption, as well as other accidents of pregnancy, were frequently associated with hypofibrinogenemia, stimulated interest in causes of intense intravascular coagulation. Although these observations were initially almost totally confined to obstetrical cases, subsequently they were made for almost all areas of medicine (Baglin, 1996). These syndromes are commonly termed consumptive coagulopathy or disseminated intravascular coagulation.

PREGNANCY HYPERCOAGULABILITY. Pregnancy normally induces appreciable increases in the concentrations of coagulation factors I (fibrinogen), VII, VIII, IX, and X. Other plasma factors and platelets do not change so remarkably. Plasminogen levels are increased considerably, yet plasmin activity antepartum is normally decreased compared with the nonpregnant state. Various stimuli act to incite the conversion of plasminogen to plasmin, and one of the most potent is activation of coagulation.

PATHOLOGICAL ACTIVATION OF COAGULATION. In normal circumstances, there is no appreciable continuous physiological intravascular coagulation. During pregnancy, there does appear to be increased activation of platelet, clotting, and fibrinolytic mechanisms in vivo. Gerbasi and colleagues (1990) found significant increases in fibrinopeptide A, b-thromboglobulin, platelet factor 4, and fibrinogen-fibrin degradation products. They concluded that this compensated, accelerated intravascular coagulation may serve for maintenance of the uteroplacental interface.

In pathological states, coagulation may be activated via the extrinsic pathway by thromboplastin from tissue destruction and perhaps via the intrinsic pathway by collagen and other tissue components when there is loss of endothelial integrity (Fig. 25-33). Tissue factor is released and complexes with factor VII. This is turn activates tenase (factor IX) and prothrombinase (factor X) complexes (Fig. 25-34). Common inciting factors in obstetrics include thromboplastin from placental abruption as well as endotoxin and exotoxins. Another mechanism is by direct activation of factor X by proteases, for example, as present in mucin or produced by neoplasms. Amnionic fluid contains abundant mucin from fetal squames, and this likely causes rapid defibrination with amnionic fluid embolism.

Consumptive coagulopathy is almost always seen as a complication of an identifiable, underlying pathological process against which treatment must be directed to reverse defibrination. Thus identification and prompt elimination of the source of the coagulopathy is the first priority in dealing with it. With pathological activation of procoagulants that triggers disseminated intravascular coagulation, there is consumption of platelets and coagulation factors in variable quantities. As a consequence, fibrin may be deposited in small vessels of virtually every organ system. Fortunately, this seldom causes organ failure. Small vessels are protected because coagulation releases fibrin monomers that combine with tissue plasminogen activator (t-PA) and plasminogen, which releases plasmin. In turn, plasmin lyses fibrinogen, fibrin monomer, and fibrin polymers to form a series of fibrinogen-fibrin derivatives. These share immunological determinants known as fibrin degradation products or split products.

SIGNIFICANCE. In addition to bleeding and circulatory obstruction, which may cause ischemia from hypoperfusion, consumptive coagulopathy may be associated with microangiopathic hemolysis. This is caused by mechanical disruption of the erythrocyte membrane within small vessels in which fibrin has been deposited. Varying degrees of hemolysis with anemia, hemoglobinemia, hemoglobinuria, and erythrocyte morphological changes are produced. This process likely causes or contributes to the hemolysis encountered with the so-called HELLP syndrome (Pritchard and colleagues, 1976).

In obstetrical syndromes involving consumptive co-agulopathy, the importance of vigorous restoration and maintenance of the circulation to treat hypovolemia and persistent intravascular coagulation cannot be overemphasized. With adequate perfusion of vital organs, activated coagulation factors and circulating fibrin and fibrin degradation products are promptly removed by the reticuloendothelial system. At the same time, hepatic and endothelial synthesis of procoagulants is promoted.

The likelihood of life-threatening hemorrhage in obstetrical situations complicated by defective coagulation will depend not only on the extent of the coagulation defects but—of great importance—on whether or not the vasculature is intact or disrupted. With gross derangement of blood coagulation, there may be fatal hemorrhage when vascular integrity is disrupted, yet no hemorrhage as long as all blood vessels remain intact.

CLINICAL AND LABORATORY EVIDENCE OF DEFECTIVE HEMOSTASIS. Bioassay is an excellent method to clinically detect or suspect significant coagulopathy. Excessive bleeding at sites of modest trauma characterizes defective hemostasis. Persistent bleeding from venipuncture sites, nicks from shaving the perineum or abdomen, trauma from insertion of a catheter, and spontaneous bleeding from the gums or nose are signs of possible coagulation defects. Purpuric areas at pressure sites may indicate incoagulable blood, or more commonly, clinically significant thrombocytopenia. A surgical procedure provides the ultimate bioassay for coagulation. Continuous generalized oozing from the skin, subcutaneous and fascial tissues, and vascular retroperitoneal space, should at least suggest coagulopathy. Such evidence also may be gained by observing continuous oozing from episiotomy incisions or perineal lacerations.

HYPOFIBRINOGENEMIA. In late pregnancy, plasma fibrinogen levels typically are 300 to 600 mg/dL. With activation of coagulation, these high levels may sometimes serve to protect against clinically significant hypofibrinogenemia. To promote clinical coagulation, fibrinogen levels must be around 150 mg/dL. If serious hypofibrinogenemia is present, the clot formed from whole blood in a glass tube may initially be soft but not necessarily remarkably reduced in volume. Then, over the next half hour or so, it becomes quite small, so that many of the erythrocytes are extruded and the volume of liquid clearly exceeds that of the clot.

FIBRIN AND FIBRINOGEN DERIVATIVES. Fibrin degradation products in serum may be detected by a number of sensitive test systems. Monoclonal antibodies to detect the D-dimer are commonly used. With clinically significant consumption coagulopathy, these measurements are always abnormally high.

THROMBOCYTOPENIA. Serious thrombocytopenia is likely if petechiae are abundant, clotted blood fails to retract over a period of an hour or so, or if platelets are rare in a stained blood smear. Confirmation is provided by platelet count.

PROTHROMBIN AND PARTIAL THROMBOPLASTIN TIMES. Prolongation of these coagulation tests may be the consequence of appreciable reductions in those coagulants essential for generating thrombin, a fibrinogen concentration below a critical level of about 100 mg/dL, or appreciable amounts of circulating fibrinogen-fibrin degradation products. Prolongation of the prothrombin time and partial thromboplastin time need not be the consequence of disseminated intravascular coagulation.

HEPARIN. The infusion of heparin to try to block disseminated intravascular coagulation associated with placental abruption or other situations in which the integrity of the vascular system is compromised is mentioned only to condemn its use.

EPSILON-AMINOCAPROIC ACID. Epsilon-aminocaproic acid has been administered to try to control fibrinolysis by inhibiting the conversion of plasminogen to plasmin and the proteolytic action of plasmin on fibrinogen, fibrin monomer, and fibrin polymer (clot). Failure to clear fibrin polymer from the microcirculation could result in organ ischemia and infarction, such as renal cortical necrosis. Its use in most types of obstetrical coagulopathy is not recommended.

ABRUPTIO PLACENTAE

Abruptio placentae is the most common cause of severe consumptive coagulopathy in obstetrics. It is discussed on page 621.

FETAL DEATH AND DELAYED DELIVERY

Although in most women with fetal death, spontaneous labor eventually ensues—most often within 2 weeks—the psychological stress imposed by carrying a dead fetus usually prompts induction of labor at the time of discovery. This also obviates the dangers of coagulation defects that may develop. Undoubtedly, the advent of more effective methods of labor induction have enhanced the desirability of early delivery (Chap. 20).

COAGULATION CHANGES. Weiner and associates reported in 1950 that some isoimmunized D-negative women who carried a dead fetus developed coagulation defects. Prospective studies indicated that gross disruption of the maternal coagulation mechanism rarely developed before less than 1 month after fetal death (Pritchard, 1959, 1973). If the fetus was retained longer, however, about 25 percent of the women developed a coagulopathy.

Typically the fibrinogen concentration falls to levels that are normal for the nonpregnant state, and in some cases it falls to potentially dangerous concentrations of 100 mg/dL or less (Pritchard, 1973). The rate of decrease commonly found is demonstrated in Figure 25-35. Simultaneously, fibrin degradation products are elevated in serum. The platelet count tends to decrease in these instances, but severe thrombocytopenia is uncommon even if the fibrinogen level is quite low. Although coagulation defects may correct spontaneously before evacuation, this is unusual and happens quite slowly (Pritchard, 1959).

PATHOGENESIS. It was clearly established that consumptive coagulopathy, presumably mediated by thromboplastin from the dead products of conception, is operational in these cases (Jimenez and Pritchard, 1968; Lerner and associates, 1967). Heparin infused alone over a few days corrected the coagulation defects, but e-aminocaproic acid did not.

HEPARIN. Correction of coagulation defects has been accomplished using low doses of heparin—5000 U two to three times daily—under carefully controlled conditions in women with an intact circulation (Weiner, 1991). Heparin appropriately administered can block further pathological consumption of fibrinogen and other clotting factors, thereby slowing or temporarily reversing the cycle of consumption and fibrinolysis. Such correction should be undertaken only if the patient is not actively bleeding, and with simultaneous steps to effect delivery.

FETAL DEATH IN MULTIFETAL PREGNANCY. It is uncommon that an obvious coagulation derangement develops in multifetal pregnancy complicated by death of at least one fetus and survival of another (Landy and Weingold, 1989). Carlson and Towers (1989) described 642 multifetal gestations of which 3 percent were complicated by death of only one fetus. Fusi and Gordon (1990) reported 11 of 485 twin pairs in which one fetus died. Santema and associates (1995) followed 29 consecutive pregnancies in which one twin died after 20 weeks. Chitkara and colleagues (1989) performed selective termination in 17 twin pairs between 20 and 24 weeks. Petersen and Nyholm (1999) followed 22 multifetal pregnancies with one fetal death after the first trimester. In none of these almost 100 cases was a coagulopathy detected.

Chescheir and Seeds (1988) reported a woman in whom, following death of one of twin fetuses, there was progressive but transient fall in plasma fibrinogen concentration and rise in fibrin split products. We have encountered only a few such cases at Parkland Hospital, and one is shown in Figure 25-36. Coagulation changes ceased spontaneously, and the surviving fetus, when delivered near term, was healthy. The placenta of the long-dead fetus was filled with fibrin. Most cases are in monochorionic twins with vascular anastomoses (Chap. 30, p. 784). The survivor-twin has an extremely high risk of cerebral palsy and other cerebral impairment (Pharoah and Adi, 2000). Although perfusion abnormalities seem the likely cause of these problems, some investigators attribute cerebral hemorrhage to disseminated intravascular coagulation and have recommended heparin administration for women with one dead twin (Romero and colleagues, 1984).

AMNIONIC FLUID EMBOLISM

This is a complex disorder classically characterized by the abrupt onset of hypotension, hypoxia, and consumptive coagulopathy. There is great individual variation in its clinical manifestation and women will be encountered in whom one of these three clinical hallmarks dominates, or is entirely absent. The syndrome is uncommon in an absolute sense; however, it is a common cause of maternal death (Berg and associates, 1996; Koonin and colleagues, 1997). Using data from 1.1 million deliveries in California, Gilbert and Danielsen (1999) estimated a frequency of about 1 case per 20,000 deliveries.

In obvious cases, the clinical picture frequently is dramatic. Classically, a woman in the late stages of labor or immediately postpartum begins gasping for air, and then rapidly suffers seizure or cardiorespiratory arrest complicated by disseminated intravascular coagulation, massive hemorrhage, and death. There appears to be great variation in clinical presentations of this condition. We have managed a number of women in whom otherwise uncomplicated vaginal delivery was followed by severe acute disseminated intravascular coagulation without cardiorespiratory symptoms. Thus, in some women, consumptive coagulopathy appears to be the forme fruste of amnionic fluid embolism (Davies, 1999; Porter and colleagues, 1996).

PATHOGENESIS. Amnionic fluid embolism was originally described in 1941 by Steiner and Luschbaugh, who found evidence of fetal debris in the pulmonary circulation of a group of women dying during labor. Subsequent studies by Adamsons and associates (1971) and Stolte and co-workers (1967), however, clearly demonstrated that amnionic fluid itself is innocuous, even when infused in large amounts. Cumulative data from the National Amniotic Fluid Embolism Registry suggested a clinical picture similar to that seen in human anaphylaxis and unlike an embolic phenomenon as commonly understood (Clark and co-workers, 1995).

Amnionic fluid enters the circulation as a result of a breach in the physiological barrier that normally exists between maternal and fetal compartments. Such events appear to be common, if not universal, with both squames of presumed fetal origin and trophoblasts being commonly found in the maternal circulation (Clark and colleagues, 1986; Lee and co-workers, 1986). There may be maternal exposure to various fetal elements during pregnancy termination, following amniocentesis or trauma, or more commonly during labor or delivery as small lacerations develop in the lower uterine segment or cervix. Alternately, cesarean section affords ample opportunity for mixture of maternal blood and fetal tissue.

In most cases, these events are innocuous. In certain women, however, such exposure initiates a complex series of physiological reactions mimicking those seen in human anaphylaxis and sepsis (Table 25-9). A similar process has been shown for traumatic fat embolism, a process once felt to involve simple vascular obstruction following trauma (Peltier, 1984). The pathophysiological cascade likely is caused by a number of chemokines and cytokines. For example, Khong (1998) found intense expression of endothelin-1 in fetal squames recovered from the lungs of two fatal cases.

PATHOPHYSIOLOGY. Studies in primates using homologous amnionic fluid injection, as well as a carefully performed study in the goat model, have provided important insights into central hemodynamic aberrations (Adamsons and co-workers, 1971; Hankins and colleagues, 1993; Stolte and colleagues, 1967). After a brief initial phase of pulmonary and systemic hypertension, there is decreased systemic vascular resistance and left ventricular stroke work index (Clark and colleagues, 1988). Transient but profound oxygen desaturation is often seen in the initial phase, resulting in neurological injury in most survivors (Harvey and associates, 1996). In women who live beyond the initial cardiovascular collapse, a secondary phase of lung injury and coagulopathy often ensues.

The association of uterine hypertonus with cardiovascular collapse appears to be the effect of amnionic fluid embolism rather than the cause (Clark and co-workers, 1995). Indeed, uterine blood flow ceases completely when intrauterine pressures exceed 35 to 40 mm Hg (Towell, 1976), thus, a hypertonic contraction is the least likely time for fetal-maternal exchange to take place. Similarly, there is no causal association between oxytocin use and amnionic fluid embolism, and the frequency of oxytocin use is not increased in these women (American College of Obstetricians and Gynecologists, 1993).

DIAGNOSIS. In the past, the detection of squamous cells or other debris of fetal origin in the central pulmonary circulation was felt to be pathognomonic for amnionic fluid embolism. Indeed, in fatal cases, histopathological findings may be dramatic, especially in those involving meconium-stained amnionic fluid (Fig. 25-37). The detection of such debris, however, may require extensive special staining and even then it is often not seen. In the National Registry, fetal elements were detected in 75 percent of autopsies and 50 percent of specimens prepared from concentrated buffy coat aspirates from a pulmonary artery catheter before death. Further, several studies have demonstrated that squamous cells, trophoblasts, and other debris of fetal origin may commonly be found in the central circulation of women with conditions other than amnionic fluid embolism. Thus, this finding is neither sensitive nor specific, and the diagnosis is generally made by identifying clinically characteristic signs and symptoms. In less typical cases, diagnosis is contingent upon careful exclusion of other causes.

MANAGEMENT. Although an initial period of systemic and pulmonary hypertension appears to be involved in amnionic fluid embolism, this phase is transient. Women who survive long enough to receive any treatment other than cardiopulmonary resuscitation should receive therapy directed at oxygenation and support of the failing myocardium. Circulatory support and blood and component replacement are paramount. There are no data that any type of intervention improves maternal prognosis with amnionic fluid embolism. In undelivered women suffering cardiac arrest, consideration should be given to emergency perimortem cesarean delivery in an effort to improve newborn outcome. However, for the mother who is hemodynamically unstable, but who has not suffered arrest, such decision making becomes more complex.

PROGNOSIS. The dismal outcomes with amnionic fluid embolism are undoubtedly related to reporting biases. Also, the syndrome likely is underdiagnosed in all but the most severe cases. In the National Registry series, there was a 60 percent maternal mortality rate. In the California database of 1.1 million deliveries by Gilbert and Danielson (1999), only a fourth of reported cases were fatal. Weiwen (2000) has provided preliminary data from 38 cases in the Suzhou region of China. Almost 90 percent of these women died. Death can be amazingly rapid, and of the 34 women who died in the series from China, 12 died with 30 minutes.

Profound neurological impairment is common in survivors. Of women reported to the National Registry who had cardiac arrest in conjunction with initial symptoms, only 8 percent survived neurologically intact. Outcome is also poor for fetuses of these latter women and is related to arrest-to-delivery interval. Overall neonatal survival is 70 percent, but almost half suffer residual neurological impairment.

SEPTICEMIA

Infections that lead to bacteremia and septic shock in obstetrics are most commonly due to septic abortion, antepartum pyelonephritis, or puerperal sepsis. Other aspects of septic shock are discussed in Chapter 43 (p. 1167).

COAGULOPATHY. The lethal properties of bacterial toxins, and especially endotoxins, are undoubtedly mediated largely by disruption of vascular endothelium. It is unclear, however, whether this is the major mechanism that initiates consumptive coagulopathy. For example, in experimental animals, endothelial damage is greatest 24 hours after endotoxin is given, but intravascular coagulation can usually be identified during the first few hours. More likely, as shown in Figure 25-34, endotoxin activates the extrinsic clotting mechanism through cytokine-induced tissue factor expression on the surface of activated monocytes (Levi and colleagues, 1993). The intrinsic route seems unimportant in this role.

MANAGEMENT. Therapy for women with septicemia from any cause is outlined in Chapter 43 (p. 1167). In general, treatment of the inciting cause will be followed by reversal of the coagulopathy. In some cases, especially if surgical procedures are performed before sepsis is controlled and the coagulopathy is reversed, treatment with fresh-frozen plasma and platelet packs usually will arrest such bleeding. Heparin therapy is dangerous and should not be given.

ABORTION

Remarkable blood loss may occur as the consequence of abortion. Hemorrhage during early pregnancy is less likely to be severe unless abortion was induced and the procedure was traumatic. When pregnancy is more advanced, the mechanisms responsible for hemorrhage are most often the same as those described for placental abruption and placenta previa—that is, the disruption of a large number of maternal blood vessels at the site of placental implantation.

COAGULATION DEFECTS. Serious disruption of the coagulation mechanism as the consequence of abortion may develop in the following circumstances:

1. Prolonged retention of a dead fetus, as already described.

2. Sepsis, a notorious cause.

3. Intrauterine instillation of hypertonic saline or urea solutions.

4. Medical induction with a prostaglandin.

5. During instrumental termination of the pregnancy.

The kinds of changes in coagulation that have been identified with abortion induced with hypertonic solutions imply at least that thromboplastin is released from placenta, fetus, decidua, or all three by the necrobiotic effect of the hypertonic solutions, which then initiates coagulation within the maternal circulation (Burkman and associates, 1977). Coagulation defects have been observed to develop rarely during induction of abortion with prostaglandin. Saraiya and colleagues (1999) reported 62 spontaneous-abortion related deaths reported to the Pregnancy Mortality Surveillance System. Almost 60 percent of deaths were caused by infection and half of these women had consumptive coagulopathy.

Consumptive coagulopathy has been an uncommon but serious complication among women with septic abortion. The incidence of coagulation defects in the past at Parkland Hospital was highest in those with Clostridium perfringens sepsis and intense intravascular hemolysis (Pritchard and Whalley, 1971). Management consists of prompt restoration and maintenance of the circulation, and appropriate steps to control the infection, including evacuation of the infected products (Chap. 33, p. 877).