Chapter  53:  Management of Prolonged Pregnancy: Evidence Report/Technology Assessment Number 53

Key Research Questions

Four key research questions were addressed:

1. What are the test characteristics (reliability, sensitivity, specificity, predictive values) and costs of measures used in the management of prolonged pregnancy (a) to assess risks to the fetus and mother of prolonged pregnancy and (b) to assess the likelihood of a successful induction of labor?

2. What is the direct evidence comparing the benefits, risks, and costs of planned induction versus expectant management at various gestational ages?

3. What are the benefits, risks, and costs of currently available interventions for the induction of labor?

4. Are the epidemiology and outcomes of prolonged pregnancy different for women in different ethnic groups, socioeconomic groups, or age groups (i.e., adolescents)?

Interventions Assessed

The following interventions were considered:

Patient Population and Settings

The primary patient population considered in the review was pregnant women with a single fetus in the vertex position, approaching or past the estimated date of confinement, without any other medical or obstetrical complications (including prior cesarean section), where the only potential factor increasing the risk of an adverse perinatal or maternal outcome was advancing gestational age. The researchers also examined the potential interaction of this risk with age and race/ethnicity. The principal practice settings considered were hospitals, freestanding birthing centers, patients' homes, and prenatal clinics or other facilities where ambulatory prenatal care is delivered.

Outcomes Considered

Outcomes considered varied depending on the study and the question being addressed, but the researchers focused primarily on clinically relevant outcomes. Data recorded included anatomic outcomes (changes in cervical dilation or Bishop score); perinatal and maternal mortality; surrogate markers of fetal compromise (nonreassuring changes in fetal heart rate patterns, meconium); mode of delivery (cesarean, vaginal, operative vaginal); other interventions (need for labor augmentation, need for labor induction); adverse outcomes (complications of vaginal and cesarean delivery, complications of interventions); and use of resources (time to delivery, length of stay, medication, and labor costs).

Literature Sources Used

The primary sources of literature were the following databases (with search years shown in parentheses) MEDLINE (1980-December 2000), HealthSTAR (1980-December 2000), CINAHL (1983-December 2000), Cochrane Database of Systematic Reviews (CDSR) (Issue 4, 2000; Issue 1, 2001; and Issue 2, 2001), Database of Abstracts of Reviews of Effectiveness (DARE), and EMBASE (1980-Jan 2000). Searches of these databases were supplemented by secondary searches of reference lists in all included articles, especially Cochrane review articles, scanning of current issues of journals not yet indexed in the computerized bibliographic databases, and suggestions from an advisory panel.

The initial searches were performed in MEDLINE and then duplicated in other databases. All searches were limited to English-language articles published since 1980 involving human subjects. The cut-off threshold of 1980 was based on the lack of general availability of ultrasound prior to that date. It was judged that trials conducted and published prior to 1980 would be problematic both in terms of the accuracy of diagnosis and comparability with current testing and management strategies. Primary MeSH terms used in all searches included "pregnancy, prolonged/" and "post$ pregnan$.tw."

Screening of Articles

The searches yielded 701 English-language articles. Abstracts from these articles were reviewed against the inclusion/exclusion criteria by six physician investigators, with assistance from one senior medical student. A team of two investigators reviewed each abstract; when no abstract was available, the title, source, and MeSH words were reviewed. At this stage, articles were included if requested by one member of the team. At the full-text screening stage, two investigators independently reviewed each article, and disagreements were resolved through discussion.

Each screened article was coded according to three topic areas: (a) testing: two or more tests were compared in terms of accuracy or agreement of test results, or the test result was correlated with some health outcome; (b) management: the article addressed the relative effectiveness of planned induction versus expectant management or the relative effectiveness of an induction agent; and (c) testing and management: some combination of the above.

Included study designs were determined by the article's topic area. Study designs for articles on testing or testing and management included randomized controlled trials, cohort studies, and large case series (at least 20 subjects). The only study design included for management articles was the randomized controlled trial.

Studies of these types were included if they met the following criteria:

• Study population included women with prolonged pregnancy.

• Study provided data relevant to at least one of the four key questions described above.

• Study reported health outcomes, use of health services, or economic outcomes related to the management of prolonged pregnancy.

Exclusion criteria included:

• Article was not original research.

• Article did not address prolonged pregnancy.

• Study design was a single case report.

• Study design was a small case series with fewer than 20 subjects.

• Article evaluated testing, but data provided were insufficient to construct 2-by-2 tables of test sensitivity and specificity.

Data Abstraction Process

Teams of two investigators performed the data abstraction for eligible articles identified at the full-text screening stage. For each included article, one physician completed the data abstraction form, and the other served as an "over-reader." The information from the data abstraction form -- including details on study characteristics, patient population, outcomes, and quality measures -- was then summarized into evidence tables. Data abstraction assignments were made based on clinical and research interests and expertise.

Criteria for Evaluating the Quality of Articles

Using criteria developed for prior evidence reports, the researchers evaluated each article for the presence or absence of factors influencing internal and external validity. These criteria were:

• For management articles: Randomized allocation to treatment and appropriate methods of randomization; adequate description of the patient population to allow comparison with the intended patient population, including descriptions in terms of gestational age, criteria used to assign gestational age, and measurement of baseline cervical ripeness; description of criteria used to make management decisions associated with primary outcomes such as cesarean delivery; and recognition and discussion of important statistical issues such as sample size and use of appropriate tests.

• For testing articles: The above criteria, plus description of an implicit or explicit reference standard, discussion of issues of verification bias, measurement of test reliability, and adequate description of the testing protocol.

Additional Data Sources

The researchers also examined discharge data from the Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample maintained by the Agency for Healthcare Research and Quality. This database contains administrative discharge data from over 1,000 hospitals in 22 States (at the time of the review), representing a stratified sample of 20 percent of U.S. hospitals. The researchers used these data to provide supplemental information on differences in the epidemiology and outcomes of prolonged pregnancy between ethnic and socioeconomic groups. Using ICD-9 codes, they divided all deliveries into "preterm" (644.2x), prolonged (645.x), and "term" (all other delivery codes). The researchers examined differences in outcomes between coded ethnic groups (white, black, Hispanic, Asian/Pacific Islander, American Indian, and other) and by insurance status (Medicare, Medicaid, private/health maintenance organization, self-pay/no insurance, "no charge," and "other") within these categories.

Menstrual Dates

Prior to the ready availability of ultrasound in the 1980s, estimation of gestational age based on menstrual dates alone was often inaccurate. For example, women who conceived soon after stopping oral contraceptives were more likely to have prolonged gestations in one series (Keng and Eng, 1982). Even with accurate recall of dates, there will be some variability in gestational age estimation because the 40-week estimate is based on an assumption of an "ideal" 28-day menstrual cycle, with ovulation on day 14. Because the follicular phase is often quite variable (ranging from 7 to 21 days), this assumption (upon which most gestational age calculators are based) will inevitably lead to some over- or underestimation of gestational age and can lead to errors in understanding the relationship between gestational age, birthweight, and pregnancy outcome (Gjessing, Skjaerven, and Wilcox, 1999).

Ultrasound

The availability of ultrasound in most sites in the United States has substantially improved the ability to estimate gestational age more precisely. Randomized trials of routine versus selective screening with ultrasound in the second trimester have consistently found a reduced incidence of induction of labor for prolonged pregnancy in the routine screening groups, presumably because of more accurate dating (Crowley, 2000). However, ultrasound itself has a nonnegligible degree of error. The error is approximately ± 1 week for scans done in the first trimester, ± 2 weeks for scans done in the second trimester, and ± 3 weeks for scans done in the third trimester (ACOG, 1997). Thus, even for women with early ultrasound dating, the "true" gestational age falls within a 14-day window of time; that is, some women with a recorded gestational age of 41 weeks will actually be 42 weeks, and some will actually be 40 weeks. In addition, because ultrasound dating is based on embryonic or fetal size, an association between size at the time of the ultrasound and later outcomes can create systematic bias in assessing gestational age-associated risk (Henriksen, Wilcox, Hedegaard, et al., 1995). For example, ultrasound dating will consistently overestimate the gestational age of larger than average fetuses. This early overestimation of gestational age could create a bias that would lead to an overestimation of the association of advanced gestational age and macrosomia. On the other hand, gestational age will be consistently underestimated for smaller than average fetuses. If some conditions that lead to low birthweight manifest themselves very early in pregnancy, then this will lead to an underestimation of the association of conditions associated with low birthweight and advancing gestational age.

The effects of uncertainty in dating pregnancy are not insignificant. Population-based estimates of the outcomes of pregnancy by gestational age, clinical trial data, and policy and clinical decisions based on these data are all dependent on the accuracy of the determination of gestational age.

The population of pregnant women with "prolonged" pregnancy thus likely represents at least two distinct groups:

1. Women in whom gestational age is overestimated because of the inherent error of all methods of dating.

2. Women whose pregnancies are correctly dated. Some of these women may represent the outer limits of normal variability. Others may have underlying defects in the mechanisms signaling the onset of labor.

It is likely that the risk of adverse outcomes varies among these groups. Many of the monitoring strategies discussed throughout this report are designed to identify fetuses at higher risk of adverse outcomes. The following section discusses the adverse outcomes associated with prolonged gestation, as well as the degree to which the risk of these outcomes is related to gestational age.

Risk of Perinatal Mortality

Table 1: Observed relationship between gestational (more...)

Table 1: Observed relationship between gestational age and stillbirth risk

Study	Location	Dates	N	Stillbirth Risk per 1,000
				Denominator: 1,000 Deliveries in Specified Gestational Week							Denominator: 1,000 Continuing Pregnancies
								37	38	39	40	41	42	43	37	38	39	40	41	42	43
Yudkin, Wood, and Redman, 1987	Oxford	1978-1985	40,888	2.14		0.43		1.24			0.42		0.29		1.24
Hilder, Costeloe, and Thilaganathan, 1998	London	1989-1991	171,527	6.2	3.8	2.2	1.5	1.7	1.9	2.1	0.35	0.56	0.57	0.86	1.27	1.55	2.12
Cotzias, Paterson-Brown, and Fisk, 1999	London	1989-1991	171,527	-	-	-	-	-	-	-	1.55	1.37	1.19	1.08	1.21	1.30	1.58
Smith, 2001	Scotland	1985-1996	700,878	-	-	-	-	-	-	-	0.4	0.4	0.5	0.9	1.2	1.9	6.3

Hilder, et al., examined data from 171,527 births from the North East Thames Region in London (Hilder, Costeloe, and Thilaganathan, 1998). Stillbirth rates calculated as a percentage of all deliveries declined from 6.2/1,000 at 37 weeks to 1.5/1,000 at 40 weeks, then began to increase again with advancing gestational age (1.7 at 41 weeks, 1.9 at 42 weeks, and 2.1 at 43 weeks or more). The pattern was slightly different when risk was estimated as stillbirths per 1,000 ongoing pregnancies: 0.34 at 37 weeks, 0.70 at 38 weeks, 0.83 at 39 weeks, 1.57 at 40 weeks, 1.48 at 41 weeks, 3.29 at 42 weeks, and 3.71 at 43 weeks and beyond.

Cotzias, Paterson-Brown, and Fisk (1999) performed a reanalysis of the data set used by Hilder's group. In addition to estimating the number of stillbirths in a given gestational age divided by the number of ongoing pregnancies, the authors also estimated the "prospective stillbirth risk," the total number of stillbirths at or beyond a given gestational age divided by the total number of pregnancies at or beyond that age, multiplied by 1,000. Other data sets were used to estimate the proportion of singleton births and the proportion of stillbirths occurring in singleton pregnancies, as well as the proportion of stillbirths that were unexplained by anomalies or other recognized fetal and maternal complications. Using this methodology, the risk for unexplained stillbirth in singleton pregnancies was highest at 37 weeks (1.55/1,000), declined to a low of 1.08/1,000 at 40 weeks, then increased again to 1.58/1,000 at 43 weeks. The high rates at lower gestational ages may reflect this methodology.

Most recently, Smith (2001) analyzed data from Scotland for the period 1985 through 1996. This analysis has several advantages over the previous ones. First, the number of deliveries is considerably larger, resulting in greater precision of risk estimates. Second, stillbirths are divided into antepartum and intrapartum stillbirths, a distinction that has clinical relevance, since clinical strategies for preventing each of these might be quite different. Third, congenital anomalies were explicitly excluded. Fourth, life table methods were used to account for censoring resulting from deliveries within a given observation period. Fifth, the time period is considerably later, making the results more likely to reflect current clinical management, at least in the United Kingdom. Finally, cumulative probabilities for stillbirth at each gestational age were estimated.

The advantage of cumulative probability is that it captures the risk of death in preceding gestational ages. Smith (2001) uses the metaphor of Russian roulette to explain the difference between conditional probability and cumulative probability: the risk with each pull of the trigger is 1 in 6, but the risk of death for someone taking his fifth shot is greater than for someone taking his first shot. For example, Smith estimated the conditional probability of stillbirth at 43 weeks as 6.3/1,000 ongoing pregnancies, while the cumulative probability was 11.5/1,000 ongoing pregnancies. This difference represents the effects of stillbirths occurring before 43 weeks. The potential clinical significance of this is that achieving the absolute minimum cumulative stillbirth probability may require interventions at earlier gestational ages.

Consistently, the risk of stillbirth in the above-described studies rises with advancing gestational age, and this increase appears to begin at 39-40 weeks when estimated using the number of ongoing pregnancies as the denominator. One limitation of these studies is that they were all performed in the United Kingdom, and the degree to which the risks would differ in a different population with different clinical management is unclear. Another limitation is that other potential causes of perinatal mortality, such as maternal diabetes or hypertension, are not explicitly accounted for in these data sets. Also, autopsy verification that fetal anomalies or other anatomic causes of death did not occur was not performed. However, a recent Norwegian case-control study of unexplained stillbirth, in which autopsy verification was performed and logistic regression was used to control for documented maternal disease, found that increasing gestational age remained a significant risk factor for unexplained stillbirth, along with maternal age, smoking, obesity, and low educational level. Interestingly, parity was not a risk factor in the multivariate analysis (Froen, Arnestad, Frey, et al., 2001).

It should be pointed out that the risk of stillbirth in these studies remains quite low at an absolute level. The point at which the risk becomes unacceptable and justifies intervention is unclear and is likely to be influenced by each couple's feelings about the tradeoffs between intervention and no intervention.

Two other studies provide additional indirect evidence of increased risk of death with prolonged gestation. Bastian, Keirse, and Lancaster (1998) compared outcomes of all planned home births in Australia from 1985 through 1990 with all Australian births in the same time period and home births in other countries. The planned home birth perinatal death rate was 6.4/1,000 (46/7,002 total home births). Of the 44 deaths with known gestational age, seven (15.9 percent) were greater than 42 weeks. On chart review, six of these deaths, or 28.6 percent of the total, were classified as due to intrapartum asphyxia; prolonged pregnancies represented 10.7 percent of all home births. Overall, the mortality rate for home births in infants over 42 weeks was twice that for other home births. The authors point out that other conditions associated with perinatal mortality are much less common in the home-birth population, so that the excess mortality observed is unlikely to be solely due to the confounding effects of other complications, such as preeclampsia or diabetes.

Mehl-Madrona and Madrona (1997) reviewed self-reported data from midwives in the western United States between 1970 and 1985. A total of 4,361 midwife-attended home births were compared with 4,107 family-practitioner-attended home births performed in California and Wisconsin during the same time period. Sampling frames and response rates were variable, as were the data collection instruments. Deliveries were matched by maternal age, insurance status, parity, and presence of risk factors. Midwives were significantly more likely to deliver postdate pregnancies, defined as gestational age greater than 42 weeks, than were family practitioners (midwives also were more likely to deliver breech and twin pregnancies). Mortality rates were significantly higher for midwives compared to family practitioners, a difference that was attributable entirely to more postdate, twin, and breech deliveries in the midwife group.

Both of these studies are limited by issues concerning accuracy of dating, completeness of reporting, confirmation of causes of death, and in the case of the Mehl-Madrona paper, a rather complicated sampling scheme and questions about the true comparability of groups. There also are concerns about generalizability in terms of current midwifery practice in the United States. However, patients who select home birth are, by definition, low-risk patients. They also are unlikely to have undergone antepartum testing. The excess mortality seen in women with prolonged pregnancy delivering at home in these two studies is consistent with an independent effect of increasing gestational age on perinatal mortality.

Causes of Perinatal Mortality in Prolonged Pregnancies

Analysis of data from the Medical Birth Registry of Norway from 1978 to 1987 found that the risk of perinatal death was over five times higher in infants below the 10^th percentile of birthweight for their gestational age (odds ratio [OR], 5.68; 95 percent confidence interval [CI], 4.37 to 7.38) than in infants from the 10^th to 90^th percentile (Campbell, Ostbye, and Irgens, 1997), after adjustment for a variety of potential confounding variables, such as maternal complications like diabetes. Maternal age > 35 years was also a risk factor in multivariate analysis (OR, 1.88; 95 percent CI, 1.22 to 2.89). Infants above the 90^th percentile in weight had a decreased mortality risk (OR, 0.51; 95 percent CI, 0.26 to 1.00). A similar relationship between perinatal mortality in prolonged pregnancy and low birthweight was found in a review of Swedish registry data from 1987 through 1992 (Divon, Haglund, Nisell, et al., 1998). These observations are consistent with a hypothesis that decreased uteroplacental function, leading to growth restriction, oligohydramnios, and eventually asphyxia, is one of the major risks of advancing gestational age, although changes in weight occurring after death and prior to delivery may explain some of this phenomenon. What is not clear is whether the decreasing uteroplacental function is an inevitable result of advancing gestational age, or whether failure to go into labor is somehow a marker for some forms of uteroplacental insufficiency.

The Norwegian data are limited by the population (results may not be generalizable to a more diverse U.S. population), accuracy of dating (gestational age in the registry is based on last menstrual period), and time (obstetric management has changed somewhat since 1987). However, the observed association between low birthweight and perinatal mortality in a genetically homogeneous population with a relatively high standard of living and level of access to prenatal care suggests that this is at least partly a reflection of changes in the biology of the uterus, placenta, and/or fetus associated with prolonged pregnancy.

Another issue that should be considered in reviewing recent population-based data on perinatal mortality is the degree to which observed perinatal deaths are preventable. It is unclear from population-based administrative data what proportion of unexplained stillbirths after 40 weeks gestation occurred in women undergoing some form of antenatal surveillance. This information is important for two reasons. First, in order to estimate the benefits of antenatal surveillance at different gestational ages quantitatively, the baseline gestational-age-specific risk, in the absence of surveillance, is needed. Second, if current mortality data reflect mostly women who are undergoing surveillance, then the limits of currently available technology may have been reached; in this case, the only strategy available for further reducing perinatal mortality would be elective induction of labor at a predefined gestational age. This is supported by the findings of a Cochrane meta-analysis (Crowley, 2000), which showed an excess of perinatal mortality in the testing arms. Conversely, if current mortality data reflect women who are not undergoing surveillance, then greater efforts are needed to ensure access to currently available technologies.

Perinatal Morbidity

In the Norwegian database, risks for fetal distress in labor (relative risk [RR], 1.68; 95 percent CI, 1.62 to 1.72) and shoulder dystocia (RR, 1.31; 95 percent CI, 1.21 to 1.42) were significantly increased in infants born after 42 weeks compared with infants born between 39 and 42 weeks (Campbell, Ostbye, and Irgens, 1997). Others also have noted an association between prolonged pregnancy and increased fetal weight and/or shoulder dystocia (Acker, Sachs, and Friedman, 1985; Eden, Seifert, Winegar, et al., 1987; Nocon, McKenzie, Thomas, et al., 1993; Sarno, Hinderstein, and Staiano, 1991).

Data on longer term outcomes of infants born after prolonged gestations are relatively sparse. One Irish case-control study reported an association between prolonged pregnancy and neonatal seizures (Curtis, Matthews, Clarke, et al., 1988). In a study of British children with cerebral palsy, there was a strong association between maternal gestational age greater than 41 weeks and the presence of neonatal encephalopathy (defined as having both signs of neonatal neurological abnormalities and depression at birth, defined as a 1-minute Apgar score less than 6) (OR, 3.5; 95 percent CI, 1.0 to 12.1). This risk was particularly marked in primigravid women (OR, 11.0; 95 percent CI, 1.5 to 102.5). The infants studied also were more likely to have had induction of labor (indications not specified), long second stage of labor, meconium-stained amniotic fluid, and emergent cesarean section or operative vaginal delivery.

On the other hand, prospective studies have not shown an association between prolonged pregnancy and adverse physical or mental development at 1 or 2 years, even when stratified by presence or absence of the dysmaturity syndrome (Shime, Librach, Gare, et al., 1986).

In summary, available data are insufficient to quantify the degree of excess risk, if any, of perinatal morbidity (including neurological morbidity) associated with prolonged pregnancy.

Maternal Outcomes

Maternal risks of obstetric trauma and hemorrhage are increased in prolonged pregnancy compared with term pregnancy (Campbell, Ostbye, and Irgens, 1997). Labor abnormalities also are increased. All three of these may be related to an increased risk of macrosomia. Another potential reason, as stated above, is that some women who do not go into labor within the "normal" length of gestation have differences in the physiology of labor and delivery compared with women who begin labor earlier in gestation.

Interventions performed to prevent adverse outcomes associated with prolonged gestation have the potential for complications, most notably hyperstimulation resulting from too frequent uterine contractions, infection, bleeding, or organ injury from cesarean section.

Summary: Risks of Prolonged Pregnancy

Prolonged gestation is associated with an increased risk of perinatal death, as well as perinatal morbidities related to either uteroplacental insufficiency or fetal macrosomia. Direct maternal risks are potentially related to fetal macrosomia or to interventions used in the management of prolonged pregnancy. The gestational age at which the risk of adverse direct perinatal or maternal outcomes justifies the costs and potential complications of active intervention is unclear.

Key Research Questions

The key research questions addressed in the report were developed by the Agency for Healthcare Research and Quality (AHRQ) and our report partner, ACOG, and refined in consultation with AHRQ, ACOG, and the project's advisory panel of technical experts. The questions were as follows:

1. What are the test characteristics (reliability, sensitivity, specificity, predictive values) and costs of measures used in the management of prolonged pregnancy to (a) assess risks to the fetus and mother of prolonged pregnancy, and (b) assess the likelihood of a successful induction of labor?

2. What is the direct evidence comparing the benefits, risks, and costs of planned induction versus expectant management at various gestational ages?

3. What are the benefits, risks, and costs of currently available interventions for the induction of labor?

4. Are the epidemiology and outcomes of prolonged pregnancy different for women in different ethnic groups, different socioeconomic groups, or in adolescent women? This question reflects AHRQ's programmatic interest in identifying health disparities attributable to age, race/ethnicity, and socioeconomic status.

Our approach to addressing each of these questions was to identify and evaluate the relevant literature and supplemental data (if any); report the results; and where evidence was lacking or methodological limitations in the available sources precluded drawing firm conclusions, identify the issues needing resolution in order to answer the question.

Because the primary focus of the report is on clinical issues surrounding advancing gestational age, we did not systematically review the basic science literature on the initiation of labor, the physiology of the gravid uterus and cervix, placental function, or any of the other topics critical to a comprehensive understanding of these issues. The Duke team, AHRQ, ACOG, and the advisory panel all agreed that the time, effort, and additional expertise required to systematically review this literature precluded their inclusion in this evidence report.

Interventions Assessed

Based on the key research questions, our preliminary review of the literature, and discussions with the advisory panel, we considered the following interventions to reduce risks to the fetus or mother associated with advancing gestational age.

• 1

  Testing:

  1. Tests to determine risk of stillbirth or compromise related to prolonged gestation:

     • Maternal measurement of fetal movement.

     • Nonstress test (NST).

     • Contraction stress test (CST), using either nipple stimulation or oxytocin.

     • Amniotic fluid measurements.

     • Biophysical profile, using either five measures (reactive NST, breathing, tone, movement, amniotic fluid) or two measures (NST, amniotic fluid).

     • Doppler measurements of umbilical or fetal cerebral blood flow.

  2. Tests to determine the risk of macrosomia.

     • Estimation of fetal weight:
       - Maternal judgment.
       - Clinical examination.
       - Ultrasound.

  3. Tests to estimate likely success of induction of labor.

     • Clinical estimation of cervical ripeness (Bishop score).

     • Fibronectin.

After discussion with the advisory panel, we did not include tests of fetal well-being that are no longer in widespread clinical use, such as estriol.

• 2

  Management options other than testing:

  • No intervention (neither induction nor testing).

  • Interventions to prevent prolonged pregnancy:
    - Scheduled sweeping of membranes.

  • Planned induction:
    - 41 weeks.
    - 42 weeks.
    - Later timing

  • Testing for fetal well-being (using tests described above):
    - Varied time of initiation (40, 41, 42 weeks).
    - Varied frequency.

• 3

  Specific agents/interventions used for the induction of labor:

  • Amniotomy.

  • Castor oil.

  • Extra-amniotic saline instillation.

  • Relaxin.

  • Sweeping of the membranes.

  • Foley catheter.

  • Nipple stimulation.

  • Oxytocin.

  • Prostaglandins:
    - Prostaglandin E₂ (gel, tablets, and inserts).
    - Misoprostol.

  • Mifepristone.

We did not systematically review certain other interventions that may play a role in managing prolonged pregnancy. Although we discuss the effect of ultrasound estimation of gestational age on the diagnosis of prolonged pregnancy above, we did not attempt to systematically review the literature on the other potential benefits, risks, and costs of routine ultrasonography in early pregnancy. Attempting to place the potential benefits of accurate gestational dating for managing advancing gestational age in the context of the other possible outcomes associated with routine ultrasound screening was well beyond the scope of the report and beyond the resources available. Similarly, we did not systematically review the literature on intrapartum interventions used in the management of common complications of prolonged pregnancy (such as oligohydramnios or meconium-stained amniotic fluid) unless identified articles clearly included data on prolonged pregnancy.

Literature Sources

The primary sources of literature were six of the most widely used computerized bibliographic databases: MEDLINE (1980-December 2000), HealthSTAR (1980-December 2000), CINAHL (1983-December 2000), the Cochrane Database of Systematic Reviews (CDSR) (Issue 4, 2000; Issue 1, 2001; and Issue 2, 2001), the Database of Abstracts of Reviews of Effectiveness (DARE), and EMBASE (1980-Jan 2000). Searches of these databases were supplemented by secondary searches of reference lists in all included articles, especially Cochrane review articles, and scanning of current issues of journals not yet indexed in the computerized bibliographic databases. Titles regularly scanned included the American Journal of Obstetrics and Gynecology, the British Medical Journal, the British Journal of Obstetrics and Gynaecology, the European Journal of Obstetrics and Gynecology and Reproductive Medicine, the International Journal of Gynecology and Obstetrics, the Journal of the American Medical Association, the Journal of Maternal-Fetal Medicine, the Journal of Obstetrics and Gynaecology, Obstetrics and Gynecology, the Lancet, and the New England Journal of Medicine. Suggestions regarding search terms and specific articles were solicited from the advisory panel during two conference calls in December 2000 and March 2001 and resulted in additions to the literature database.

Search Strategy

We developed the basic search strategies using the National Library of Medicine's MeSH key word nomenclature developed for MEDLINE. The same strategies were used to search HealthSTAR and CINAHL. A Duke University Medical Center librarian checked the strategies and assisted with their translation to the key word structure used by EMBASE. Dr. Evan Myers searched the CDSR and DARE using "postterm pregnancy," "prolonged pregnancy," and similar terms.

The initial searches were performed in MEDLINE and then duplicated in other databases. All searches were limited to articles published since 1980, in the English language, and with human subjects. The cut-off threshold of 1980 was based on the general unavailability of ultrasound prior to that date. It was judged that trials conducted and published prior to 1980 would be problematic both in terms of the accuracy of diagnosis and comparability with current testing and management strategies. The decision to restrict the literature search to articles published since 1980 was agreed to by the members of the advisory panel.

Table 2: Original search strategy

Table 2: Original search strategy

fetal death/ stillbirth.tw. exp fetal monitoring/ exp labor, induced/ exp cesarean section/ exp labor complications/ labor complications/ amniotom$.tw. oxytocin/ misoprostol/ dinoprost/ dinoprostone/ carboprost/ mifepristone/ fetal macrosomia/ or/1-15 pregnancy, prolonged/ post$ pregnan$.tw. or/17-18 and/16,19 limit 20 to human limit 21 to english language limit 22 to yr=1977-2001 limit 23 to abstracts limit 22 to yr=1980-2001 limit 25 to abstracts

Table 3: Additional search terms suggested by advisory (more...)

Table 3: Additional search terms suggested by advisory panel

nipple stimulation.mp. exp ACUPUNCTURE/or acupuncture.mp. exp Transcutaneous Electric Nerve Stimulation/or transcutaneous nerve stimulation.mp. exp Castor Oil/or castor oil.mp. exp HERBS/or herb.mp. PROSTAGLANDINS F, SYNTHETIC/or exp PROSTAGLANDINS/or prostaglandins.mp. or PROSTAGLANDINS E/or PROSTAGLANDINS E, SYNTHETIC/or PROSTAGLANDINS F/ or/1-6 pregnancy, prolonged/ post$ pregnan$.tw. 8 or 9 7 and 10 limit 11 to (human and english language and yr=1980-2001) (44) from 12 keep 1-44 (44) limit 13 to english language (44) limit 14 to human (44) limit 15 to yr=1980-2001 (44)

Screening Criteria

Inclusion and exclusion criteria were developed for the literature searches so that the yield of articles would be appropriately focused. Empirical studies or review articles were excluded after screening based on the following criteria:

• Article was not original research.

• Article did not address prolonged pregnancy.

• The study design was a single case report.

• The study design was a small case series with fewer than 20 subjects.

Each screened article was coded as addressing one of three topic areas:

1. Testing: Two or more tests were compared in terms of the accuracy or agreement of test results or the test result was correlated with some health outcome.

2. Management: The article addressed the relative effectiveness of planned induction versus expectant management or the relative effectiveness of an induction agent.

3. Testing and management: Some combination of the above.

The criteria used to include articles were:

• The study population must address prolonged pregnancy; ideally, results should be reported separately for patients with prolonged pregnancy. Because it is possible that the response of the cervix and uterus to induction agents would be quite different in different clinical scenarios (both in terms of labor patterns and potential maternal and fetal side effects), studies of induction agents that did not include any otherwise healthy women with prolonged pregnancy were excluded.

• All original research or relevant reviews must relate to at least one of the four key questions described above.

• Outcomes were included if they were health outcomes or health services use or economic outcomes related to the management of prolonged pregnancy.

• We included only randomized controlled trials (RCTs) which used active or nonactive (i.e., placebo) controls for studies involving management topics. For testing articles, we included RCTs and those cohort and large case series that allowed construction of 2-by-2 tables for estimation of sensitivity and specificity. Articles that did not meet these criteria were not necessarily excluded from the review and often provided valuable background material. However, only articles meeting the inclusion criteria were formally abstracted into evidence tables.

Included study designs were determined by the article's topic area. Study designs initially included for testing articles and testing and management articles were case reports; small case series (< 20 subjects); medium to large case series (> 20 subjects); nonrandomized comparison studies (cohort or case series that used historical or concomitant nonrandomized controls); and RCTs. The study design of each screened article was coded in our literature database.

Screening Results

Table 4: Number of articles identified, by literature (more...)

Table 4: Number of articles identified, by literature database

Database	Number of unduplicated articles	Percent of total	Number of articles included	Percent of included articles
MEDLINE	458	65.3	257	56.1
EMBASE	162	23.1	60	37.0
CINAHL, HealthSTAR	22	3.1	9	40.9
Cochrane Database of Systematic Reviews, DARE	25	3.6	15	60.0
Other (e.g., manual screening of references lists)	34	4.9	23	67.6
Total	701	100.0	364	--

At the full-text screening stage, each article was independently reviewed by two investigators, who forwarded their decisions to Ms. Jane Kolimaga, the task order manager, for recording and comparison. If indicated, reviewers were asked to reconcile differences of opinion. Overall, the teams initially disagreed on about 25-35 percent of their decisions, and all disagreements were resolved by consensus. In the event that two investigators could not agree, Dr. Evan Myers, the principal investigator, was to be the arbiter, but this situation never arose.

Table 5: Results of abstract screening and full-text (more...)

Table 5: Results of abstract screening and full-text article reviews

Description	No. of Unduplicated Articles Identified	Percent
Abstracts Number of titles and abstracts reviewed	701
Number of abstracts included:    Management    Testing    Testing & Management	511 317 173 21	72.9
Number of abstracts excluded:    Management    Testing    Testing & Management	190 169 11 10	27.1
Full-Text Articles Number of full-text articles reviewed	511
Number of full-text articles included:    Management    Testing    Testing & Management	364 231 107 26	71.2
Classification of included articles:    Management-RCT    Management-Review    Management-Other (e.g., epidemiology)    Management-Basic Science    Management-Background    Testing-Case Report    Testing-Small Case Series    Testing-Large Case Series    Testing-Non-Randomized Comparison    Testing-RCT    Testing-Review    Testing-Other    Testing-Basic Science    Testing-Background    Testing & Management -Large Case Series    Testing & Management -Non-Randomized Comparison    Testing & Management -RCT    Testing & Management -Review    Testing & Management -Background	85 41 41 10 54 0 5 36 0 0 8 3 5 50 3 0 4 16 3
Number of full-text articles excluded:    Management    Testing    Testing & Management    Other (e.g., not English, inaccurate reference)	147 80 60 4 3	28.8

Assessment of Risks to Fetus and Mother

In Chapter 1, we discussed the evidence for increasing risk of adverse outcomes, especially perinatal death, as gestational age advances beyond 40 weeks. Although this risk is small in absolute terms, the trend towards increasing risk with increasing gestational age is consistent across studies. One approach to preventing these adverse outcomes would be to use testing to identify patients most likely to experience them.

However, most of the published literature consists of case series or cohort studies in which there is little or no variation in testing strategies (or variation is not reported). Such studies are less useful but still may contain valuable information concerning the association of test results with fetal and maternal outcomes.

Table 6: Number of articles providing sensitivity and (more...)

Table 6: Number of articles providing sensitivity and specificity data, by test studied

Test	Number of articles
Amniotic fluid volume	11
Nonstress test	11
Fetal weight estimation by ultrasound	5
Biophysical profile - simple	4
Biophysical profile - complex	2
Doppler umbilical artery flow	2
Fetal breathing movements	2
Fetal heart rate (bradycardia, tachycardia, variation)	2
Fibronectin	2
Amniocentesis (meconium staining)	1
Cervical exam	1
Combination	1
Contraction stress test	1
Fetal motion on ultrasound	1
Fetal weight estimation - clinical exam	1
Kick counts	1

Reliability of Tests

We additionally sought data on the reliability of tests, including interobserver variation, when these were available. If a test result is not reproducible when the test is performed by different examiners, or by the same examiner on different occasions, then the utility of the test is reduced, even if the "average" test characteristics (sensitivity, specificity) imply useful discrimination or prediction.

Correlation of Tests

In certain cases, the association of one test result with another was reported without reference to outcomes.

Assessment of Risks to the Fetus Associated with Uteroplacental Insufficiency

Maternal sensation of fetal movement (kick counts)

We identified only one study that assessed the association of maternal sensation of fetal movement with postmaturity syndrome, defined as characteristic skin changes (desquamation, leather-like consistency, little subcutaneous fat) and a "long, lean body," with a ponderal index (weight in grams x 100/length in cubic centimeters) of 2.27 or less (10^th percentile or less). Rayburn, et al., tested a group of 147 women at 42 weeks or more gestational age using the NST plus fetal movement charting plus urine estrogen-to-creatinine ratio (Rayburn, Motley, Stempel, et al., 1982). These tests were performed semi-weekly or weekly. If the NST was reactive (two adequate accelerations of baseline fetal heart rate [FHR] during a 20- to 40-minute period), then it was repeated on the next visit. If the NST was nonreactive, then the test was either repeated or a contraction stress test (CST) was given on the same day. Of the 147 cases studied, 32, or 22 percent, had postmaturity syndrome. However, none of the mothers recording kick counts noted reduced fetal movement (sensitivity, 0/32; specificity, 115/115 [100 percent]). The kick count measure was not useful for predicting postmaturity syndrome, with an undefined positive predictive value and negative predictive value of 78 percent. No studies documenting the reliability of this method (such as correlation between maternal sensation of movement and observed movements on ultrasound) were identified.

In summary, there are no data to suggest that maternal sensation of fetal movement is useful in predicting which infants are affected by postmaturity syndrome. There are no data at all to allow evaluation of maternal sensation of fetal movement as a predictor of other adverse outcomes associated with prolonged gestation.

Nonstress test (NST)

We identified one randomized trial enrolling 287 patients comparing the NST alone with a simple biophysical profile (NST plus AFV, supplemented by estimates of fetal weight and placental function) (Arias, 1987). In this trial, 44 of 217 patients had abnormal results on antenatal testing, 14/112 in the NST alone group and 30/105 in the NST + AFV group. There were no significant differences in any outcome, including fetal distress or cesarean section for fetal distress, though slightly more inductions and cesarean sections for fetal distress occurred in the biophysical profile arm. Test characteristics of other components of this combination of tests (ultrasound for fetal weight alone, ultrasound for placental function alone, or ultrasound for AFV alone) were not reported. Sensitivity was similar for NST alone and NST + AFV; however, specificity was higher for NST alone than for NST + AFV. This study was rated positively for 9 of 12 quality assessment items, failing items for sample size and statistical analysis.

Table 7: Outcomes reported in association with test (more...)

Table 7: Outcomes reported in association with test results in 2-by-2 tables

Outcome	Number of 2-by-2 Tables
1-minute Apgar score	6
5-minute Apgar score	7
Any	2
Cesarean section for fetal distress	5
Fetal distress	3
Fetal growth restriction	1
Macrosomia	2
Meconium aspiration	2
Meconium staining	1
Neonatal mortality	3
NICU Admission	3
pH	1
Postmaturity syndrome	2
Stillbirth	2

NICU = Neonatal intensive care unit

Table 8: Nonstress test performance characteristics (more...)

Table 8: Nonstress test performance characteristics

Study	Result	Rate of Abnormal Tests	Outcome	Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	1-min Apgar	< 7	0.07	0.40	0.85	0.16	0.95
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	1-min Apgar	< 7	0.07	0.27	0.97	0.40	0.95
Tongsong and Srisomboon, 1993	Abnormal	0.20	1-min Apgar	< 7	0.07	0.41	0.81	0.14	0.95
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.13	1-min Apgar	< 7	0.15	0.42	0.92	0.50	0.90
Small, Phelan, Smith, et al., 1987	Abnormal	0.11	1-min Apgar	< 7	0.18	0.13	0.90	0.22	0.82
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	1-min Apgar	< 7	0.20	0.23	0.89	0.34	0.83
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	5-min Apgar	< 7	0.03	0.29	0.84	0.05	0.97
Tongsong and Srisomboon, 1993	Abnormal	0.20	5-min Apgar	< 7	0.02	0.33	0.80	0.04	0.98
Small, Phelan, Smith, et al., 1987	Abnormal	0.11	5-min Apgar	< 7	0.02	0.11	0.89	0.02	0.98
Devoe and Sholl, 1983	Abnormal	0.16	5-min Apgar	< 7	0.03	0.00	0.84	0.00	0.97
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.13	5-min Apgar	< 7	0.10	0.50	0.91	0.40	0.94
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	5-min Apgar	< 7	0.03	0.33	0.87	0.06	0.98
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	5-min Apgar	< 7	0.03	0.14	0.96	0.10	0.97
Weiner, Reichler, Zlozover, et al., 1993	Abnormal	0.00	Any	Yes	0.08	0.08	0.95	0.14	0.92
Arias, 1987	Abnormal	0.13	Any	Yes	0.16	0.33	0.91	0.43	0.88
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	C-section for FD	Yes	0.11	0.42	0.87	0.29	0.92
Small, Phelan, Smith, et al., 1987	Abnormal	0.11	C-section for FD	Yes	0.04	0.21	0.90	0.08	0.96
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	C-section for FD	Yes	0.05	0.31	0.88	0.13	0.96
Farmakides, Schulman, Winter, et al., 1988	Abnormal	0.43	C-section for FD	Yes	0.28	0.67	0.66	0.43	0.84
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	C-section for FD	Yes	0.11	0.23	0.98	0.60	0.91
Tongsong and Srisomboon, 1993	Abnormal	0.20	Fetal distress	Yes	0.04	0.64	0.82	0.14	0.98
Weiner, Farmakides, Schulman, et al., 1994	Abnormal	0.00	Fetal distress	Yes	0.12	0.28	1.00	0.92	0.91
Devoe and Sholl, 1983	Abnormal	0.16	Fetal distress	Yes	0.17	0.28	0.87	0.31	0.85
Farmakides, Schulman, Winter, et al., 1988	Abnormal	0.43	FGR	Not defined	0.11	0.60	0.59	0.15	0.93
Small, Phelan, Smith, et al., 1987	Abnormal	0.11	Macrosomia	> 4000 g	0.21	0.05	0.88	0.10	0.78
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	Macrosomia	> 4000 g	0.22	0.08	0.85	0.13	0.77
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.13	Meconium aspiration	Yes	0.08	0.17	0.88	0.10	0.93
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	Meconium aspiration	Yes	0.08	0.26	0.88	0.16	0.93
Ramrekersingh-White, Farkas, Chard, et al., 1993	Abnormal	0.10	Meconium staining	Yes	0.09	0.33	0.93	0.31	0.93
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	Neonatal mortality	Yes	0.01	1.00	0.97	0.30	1.00
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	Neonatal mortality	Yes	0.01	1.00	0.84	0.08	1.00
Weiner, Farmakides, Schulman, et al., 1994	Abnormal	0.04	Neonatal mortality	Yes	0.06	0.50	0.97	0.08	1.00
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	NICU Admission	Yes	0.03	0.57	0.97	0.40	0.99
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	NICU Admission	Yes	0.03	0.57	0.85	0.11	0.98
Farmakides, Schulman, Winter, et al., 1988	Abnormal	0.43	NICU Admission	Yes	0.17	0.54	0.59	0.22	0.86
Weiner, Farmakides, Schulman, et al., 1994	Abnormal	0.04	pH	< 7.2	0.03	0.70	0.98	0.58	0.99
Small, Phelan, Smith, et al., 1987	Abnormal	0.11	Postmaturity syndrome	Yes	0.07	0.13	0.89	0.08	0.93
Phelan, Platt, Yeh, et al., 1984	Abnormal	0.13	Postmaturity syndrome	Yes	0.17	0.10	0.86	0.13	0.83
Fleischer, Schulman, Farmakides, et al., 1985	Inconclusive	0.17	Stillbirth	Yes	0.09	0.00	0.83	0.00	0.99
Fleischer, Schulman, Farmakides, et al., 1985	Abnormal	0.04	Stillbirth	Yes	0.09	0.00	0.96	0.00	0.99

C-section = Cesarean section

FD = Fetal distress

FGR = Fetal growth restriction

NICU = Neonatal intensive care unit

Table 9: Specific fetal heart rate abnormalities on (more...)

Table 9: Specific fetal heart rate abnormalities on antepartum testing

Study	Screening Test Threshold	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Rayburn, Motley, Stempel, et al., 1982	Abnormal	0.02	Postmaturity syndrome	Yes	0.22	0.09	1.00	1.00	0.80
Sherer, Onyeije, Binder, et al., 1998	Bradycardia	0.26	5-min Apgar	< 7	0.23	0.21	0.72	0.18	0.75
Sherer, Onyeije, Binder, et al., 1998	Bradycardia	0.26	Meconium aspiration	Yes	0.06	0.25	0.74	0.05	0.94
Sherer, Onyeije, Binder, et al., 1998	Bradycardia	0.26	NICU Admission	Yes	0.11	0.35	0.75	0.14	0.91
Sherer, Onyeije, Binder, et al., 1998	Tachycardia	0.32	5-min Apgar	< 7	0.18	0.41	0.70	0.23	0.85
Sherer, Onyeije, Binder, et al., 1998	Tachycardia	0.32	Meconium aspiration	Yes	0.03	0.67	0.69	0.06	0.98
Sherer, Onyeije, Binder, et al., 1998	Tachycardia	0.32	NICU Admission	Yes	0.05	0.40	0.68	0.06	0.95

NICU = Neonatal intensive care unit

In summary, results of these studies suggest that a reactive nonstress test in prolonged pregnancy has good negative predictive value -- i.e., adverse outcomes are unlikely to occur in the setting of a reactive nonstress test -- but that the positive predictive values are low. Data from the one randomized trial comparing weekly NST beginning beyond 40 weeks to NST and amniotic fluid assessment suggest equivalent outcomes.

Contraction stress test (CST) using oxytocin

Knox, et al., compared the CST using oxytocin with amniocentesis for meconium staining in 187 women at 42 weeks gestation (Knox, Huddleston, and Flowers, 1979). The study was prospective, with women assigned to groups according to the last digit of hospital number. Amniocentesis was obtained on all women at entry into the study, and labor was induced immediately if meconium staining was observed. If no meconium staining was present on initial amniocentesis, then subsequent monitoring was as follows: women in the amniocentesis group received weekly amniocentesis and were induced if meconium staining was present; and women in the CST group received an immediate CST, repeated weekly if normal. Labor was induced in significantly more women in the amniocentesis group than the CST group (11/90 [12 percent] vs. 29/90 [2 percent], respectively; p < 0.005). There were no statistically significant differences between testing groups for any outcome, including Apgar score < 7 at 1 minute, Apgar score < 7 at 5 minutes, low birthweight (< 10^th percentile), neonatal morbidity, perinatal death, cesarean sections, or abnormal labor (prolonged latent phase, primary dysfunctional labor, secondary arrest of dilatation, or arrest). However, the proportion of babies with Apgar scores less than 7 at 1 and 5 minutes was two-fold higher in the amniocentesis group; the study may have been underpowered to detect this difference.

Table 10: Test performance characteristics of contraction (more...)

Table 10: Test performance characteristics of contraction stress testing using oxytocin

Study	Screening Test Threshold	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Devoe and Sholl, 1983	Abnormal	0.09	Fetal distress	Yes	0.17	0.14	0.92	0.27	0.84
Devoe and Sholl, 1983	Abnormal	0.09	5-min Apgar	<7	0.03	0.00	0.91	0.00	0.97

In summary, CST is at least equivalent to amniocentesis for meconium staining in terms of outcomes, with significantly fewer inductions; perhaps on the basis of this trial, amniocentesis is no longer used for this indication. In the setting of prolonged pregnancy, CST, when used sequentially for followup of abnormal NST, has good negative predictive value but poor positive predictive value, based on one observational study.

CST using nipple stimulation

We did not identify any studies where nipple stimulation was the sole method for performing contraction stress tests in the management of prolonged pregnancy.

Amniotic fluid measurements

We identified one relevant randomized trial. Alfirevic, et al., compared two ultrasonographic measurements of oligohydramnios, namely amniotic fluid index (AFI) < 7.3 and maximum pool depth (MPD) < 2.1 cm, among 500 women at greater than 40 weeks gestation (Alfirevic, Luckas, Walkinshaw, et al., 1997). Both groups also had NST every 3 days. There were no differences in fetal outcomes between the two strategies; however, abnormal NST was more often an indication for induction in the AFI group than in the MPD group (15 percent vs. 8 percent; p = 0.04). The overall rates of induction of labor were not statistically different between groups (87/250 vs. 77/250; p = 0.39). There was a trend toward cesarean section for fetal distress being more common in the AFI group than in the MPD group (8 percent vs. 4 percent; p = 0.09). One possible explanation for this is a lower threshold for a diagnosis of fetal distress or for performing cesarean section in the presence of nonreassuring fetal heart rate tracings or abnormal antepartum NST results. Since such results were more common in the AFI group, it is not surprising that cesareans for fetal distress also were more common.

In a comparative cohort study, Eden, et al., reported a series of 585 patients managed in one of three ways (based on temporal changes in the protocol used): (1) weekly NST with CST for nonreactive NST (from November 1, 1978 through August 31, 1979); (2) semi-weekly NST with biophysical profile for nonreactive NST (from September 1, 1979 through December 31, 1980); or (3) semi-weekly NST with biophysical profile for nonreactive NST, plus weekly AFV measurement (from January 1, 1981 through August 31, 1981) (Eden, Gergely, Schifrin, et al., 1982). The groups employing the biophysical profile had lower incidences of low Apgar score at 5 minutes, meconium aspiration, stillbirth, fetal distress requiring intervention (persistent abnormal FHR patterns), and morbidity (defined as presence of any of following: fetal distress requiring intervention, 5-minute Apgar score < 7, neonatal resuscitation, postmaturity syndrome, or meconium aspiration). However, the rate of cesarean sections was significantly higher in the groups using the biophysical profile than in the group using NST + CST alone (NST + CST, 11.5 percent; NST + biophysical profile, 29.9 percent; NST + AFV + biophysical profile, 29.4 percent; 1 vs. 2, p < 0.05; 1 vs. 3, p < 0.05). This suggests that tests using the biophysical profile may be more sensitive at identifying fetuses at risk, but that subsequent induction resulted in higher cesarean section rates. Alternatively, as discussed above, physician thresholds for performing cesarean section may be quite different based on knowledge of antepartum test results. Despite the higher rates of cesarean section, the incidence of fetal distress requiring intervention was substantially lower in the groups using biophysical profile testing in addition to NST (NST + CST, 21.8 percent; NST + biophysical profile, 4.5 percent; NST + AFV + biophysical profile, 5.5 percent; 1 vs. 2, p < 0.05; 1 vs. 3, p < 0.05).

Tongsong and Srisomboon (1993) performed NST and AFV in 242 women at 42 weeks or more in gestational age. AFV was more accurate than NST in predicting intrapartum fetal distress (p < 0.05) (AFV: sensitivity, 73 percent; specificity, 91 percent; positive predictive value, 27 percent; negative predictive value, 99 percent; NST: sensitivity, 64 percent; specificity, 82 percent; positive predictive value, 14 percent; negative predictive value, 98 percent). Given that the definition of intrapartum fetal distress included moderate to severe variable decelerations, which would be more likely in a setting of oligohydramnios, which in turn would be more likely to be detected with ultrasound, these results are not surprising.

Table 11: Amniotic fluid volume measurement for oligohydramnios (more...)

Table 11: Amniotic fluid volume measurement for oligohydramnios based on various criteria

Study	Screening Test Threshold	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Montan and Malcus, 1995	< 5 cm	0.13	1-min Apgar	< 7	0.03	0.00	0.87	0.00	0.96
Tongsong and Srisomboon, 1993	Abnormal	0.12	1-min Apgar	< 7	0.07	0.47	0.91	0.27	0.96
O'Reilly-Green and Divon, 1996	< 5 cm	0.11	1-min Apgar	< 8	0.15	0.08	0.88	0.10	0.85
Phelan, Platt, Yeh, et al., 1985	< 1 cm	0.03	1-min Apgar	< 7	0.21	0.12	0.99	0.86	0.81
Phelan, Platt, Yeh, et al., 1985	Decreased (subjective)	0.19	1-min Apgar	< 7	0.21	0.37	0.86	0.40	0.84
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.28	1-min Apgar	< 7	0.11	0.52	0.75	0.21	0.93
O'Reilly-Green and Divon, 1996	< 5 cm	0.11	5-min Apgar	< 9	0.05	0.08	0.89	0.04	0.94
Montan and Malcus, 1995	< 5 cm	0.13	5-min Apgar	< 7	0.02	0.00	0.87	0.00	0.97
Tongsong and Srisomboon, 1993	Abnormal	0.12	5-min Apgar	< 7	0.02	0.33	0.89	0.07	0.98
Sarkar and Duthie, 1997	Decreased	0.10	5-min Apgar	< 7	0.05	0.00	0.90	0.00	0.95
Phelan, Platt, Yeh, et al., 1985	< 1 cm	0.03	5-min Apgar	< 7	0.03	0.25	0.98	0.29	0.97
Phelan, Platt, Yeh, et al., 1985	Decreased (subjective)	0.19	5-min Apgar	< 7	0.03	0.38	0.82	0.07	0.97
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.28	5-min Apgar	< 7	0.04	0.88	0.74	0.13	0.99
O'Reilly-Green and Divon, 1996	< 5 cm	0.11	Any	Yes	0.06	0.14	0.89	0.08	0.94
Leveno, Quirk, Cunningham, et al., 1984	Abnormal	0.39	Any	Yes	0.01	1.00	0.61	0.02	1.00
Weiner, Reichler, Zlozover, et al., 1993	Decreased	0.08	Any	Yes	0.08	0.25	0.94	0.27	0.93
Montan and Malcus, 1995	< 5 cm	0.13	C-section	Yes	0.14	0.08	0.87	0.09	0.86
Tongsong and Srisomboon, 1993	< 3 cm	0.12	C-section for FD	Yes	0.04	0.73	0.91	0.27	0.99
Leveno, Quirk, Cunningham, et al., 1984	Abnormal	0.39	C-section for FD	Yes	0.08	0.61	0.63	0.13	0.95
Phelan, Platt, Yeh, et al., 1985	Oligohydramnios	0.03	C-section for FD	Yes	0.06	0.23	0.98	0.43	0.96
Phelan, Platt, Yeh, et al., 1985	Decreased (subjective)	0.19	C-section for FD	Yes	0.06	0.69	0.84	0.20	0.98
Sarkar and Duthie, 1997	Decreased	0.10	C-section for FD	Yes	0.20	0.17	0.92	0.33	0.82
Monaghan, O'Herlihy, and Boylan, 1987	Abnormal	0.16	C-section for FD	Yes	0.02	0.33	0.84	0.03	0.99
Crowley, O'Herlihy, and Boylan, 1984	< 3 cm	0.19	C-section for FD	Yes	0.03	0.78	0.82	0.11	0.99
Leveno, Quirk, Cunningham, et al., 1984	Abnormal	0.39	FGR	Yes	0.08	0.50	0.61	0.10	0.94
Monaghan, O'Herlihy, and Boylan, 1987	Abnormal	0.16	FGR	< 10th centile	0.12	0.48	0.88	0.34	0.93
Sarkar and Duthie, 1997	Decreased	0.10	FGR	Yes	0.01	1.00	0.91	0.11	1.00
Crowley, O'Herlihy, and Boylan, 1984	< 3 cm	0.19	FGR	< 10th centile	0.11	0.46	0.84	0.26	0.93
Sarkar and Duthie, 1997	Decreased	0.10	Intubation	Yes	0.03	0.00	0.90	0.00	0.97
Phelan, Platt, Yeh, et al., 1985	Decreased (subjective)	0.19	Macrosomia	> 4000 g	0.22	0.12	0.79	0.13	0.76
Phelan, Platt, Yeh, et al., 1985	< 1 cm	0.03	Macrosomia	> 4000 g	0.22	0.00	0.96	0.00	0.77
Eden, Gergely, Schifrin, et al., 1982	Abnormal	0.28	Meconium aspiration	Yes	0.03	0.80	0.73	0.08	0.99
Monaghan, O'Herlihy, and Boylan, 1987	Abnormal	0.16	Neonatal mortality	Yes	0.01	0.00	0.84	0.00	0.99
Sarkar and Duthie, 1997	Decreased	0.10	NICU admission	Yes	0.01	1.00	0.91	0.06	1.00
Monaghan, O'Herlihy, and Boylan, 1987	Abnormal	0.16	NICU admission	Yes	0.09	0.17	0.84	0.09	0.91
Crowley, O'Herlihy, and Boylan, 1984	< 3 cm	0.19	NICU admission	Yes	0.07	0.38	0.82	0.14	0.94
Monaghan, O'Herlihy, and Boylan, 1987	Abnormal	0.16	pH	< 7.25	0.07	0.23	0.84	0.09	0.94
Rayburn, Motley, Stempel, et al., 1982	Oligohydramnios	0.56	Postmaturity syndrome	Yes	0.22	0.91	0.54	0.35	0.95
Rayburn, Motley, Stempel, et al., 1982	Pockets	0.20	Postmaturity syndrome	Yes	0.22	0.75	0.96	0.83	0.93

C-section = Cesarean section; FD = Fetal distress; FGR = Fetal growth restriction; NICU = Neonatal intensive care unit

Abdominal palpation

As part of an investigation of the value of ultrasound evaluation of amniotic fluid volume in predicting adverse outcomes, Crowley, et al., also evaluated the performance of clinical assessment of AFV by abdominal palpation. This technique had a false positive rate of 25 percent and a false negative rate of 43 percent for predicting "significant meconium staining or absent amniotic fluid" at the time of amniotomy (Crowley, O'Herlihy, and Boylan, 1984).

Simple biophysical profile

Table 12: Components of biophysical profile scores (more...)

Table 12: Components of biophysical profile scores in included studies

Study	NST	Amniotic Fluid Measurement		Fetal Breathing Movements	Fetal Tone/Movements	Fetal Body Measurements	Uterine Artery Resistance	Fetal Dopplers	Fetal Reflexes (Magnitude and Speed)	Placental Grading
		Maximum Pool Depth	Amniotic Fluid Index
Alfirervic and Walkinshaw, 1995: "Simple"	X	X
Alfirervic and Walkinshaw, 1995: "Complex"	X		X	X	X	X
Arias, 1987	X	X				X				X
Bochner, Medearis, Ross, et al., 1987	X	X			X
Bochner, Williams, Castro, et al., 1988	X	X			X
Brar, Horenstein, Medearis, et al., 1989	X		X				X	Umbilical
Eden, Gergely, Schifrin, et al., 1982: "modified biophysical profile"	X	X		X	X
Arabin, Snyjders, Mohnhaupt, et al., 1993: "traditional biophysical profile"	X	X		X	X
Arabin, Snyjders, Mohnhaupt, et al., 1993: "fetal assessment score"	X				X		X	Carotid	X
Hann, McArdle, and Sachs, 1987	X	X		X	X					X
Gilson, O'Brien, Vera, et al., 1988	X	X		X	X

NST = Nonstress test

Table 13: Simple biophysical profile test performance (more...)

Table 13: Simple biophysical profile test performance characteristics

Study	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Bochner, Medearis, Ross, et al., 1987	0.23	1-min Apgar	< 7	0.39	0.25	0.79	0.43	0.63
Bochner, Medearis, Ross, et al., 1987	0.23	5-min Apgar	< 7	0.02	1.00	0.79	0.07	1.00
Brar, Horenstein, Medearis, et al., 1989	0.42	5-min Apgar	< 7	0.18	0.88	0.68	0.37	0.96
Arias, 1987	0.29	Any	Yes	0.39	0.37	0.77	0.50	0.65
Bochner, Medearis, Ross, et al., 1987	0.23	C-section for FD	Yes	0.21	0.85	0.94	0.79	0.96
Brar, Horenstein, Medearis, et al., 1989	0.42	C-section for FD	Yes	0.29	0.69	0.69	0.47	0.85
Bochner, Williams, Castro, et al., 1988	0.22	Fetal distress	Yes	0.06	0.81	0.82	0.22	0.99
Bochner, Williams, Castro, et al., 1988	0.15	Fetal distress	Yes	0.03	0.70	0.87	0.12	0.99
Bochner, Medearis, Ross, et al., 1987	0.23	FGR	< 10th percentile	0.11	0.29	0.78	0.14	0.90
Bochner, Medearis, Ross, et al., 1987	0.23	Meconium aspiration	Yes	0.05	0.33	0.78	0.07	0.96
Bochner, Medearis, Ross, et al., 1987	0.23	Neonatal mortality	Yes	0.00		0.77	0.00	1.00
Brar, Horenstein, Medearis, et al., 1989	0.42	NICU Admission	Yes	0.13	0.83	0.64	0.26	0.96

C-section = Cesarean section

FD = Fetal distress

FGR = Fetal growth restriction

NICU = Neonatal intensive care unit

Complex biophysical profile score

Table 14: Complex biophysical profile test performance (more...)

Table 14: Complex biophysical profile test performance characteristics

Study	Test Threshold	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Arabin, Snyjders, Mohnhaupt, et al., 1993	Multiple	Varies	C-section for fetal distress	Yes	0.35	NR	NR	NR	NR
Arabin, Snyjders, Mohnhaupt, et al., 1993	Multiple	Varies	1-min Apgar	< 7	0.09	NR	NR	NR	NR
Arabin, Snyjders, Mohnhaupt, et al., 1993	Multiple	Varies	PH	Yes	0.08	NR	NR	NR	NR
Gilson, O'Brien, Vera, et al., 1988	< 8	0.11	1-min Apgar	< 7	0.20	0.08	0.88	0.14	0.79
Gilson, O'Brien, Vera, et al., 1988	< 8	0.20	5-min Apgar	< 7	0	Undefined	0.80	0	1
Hann, McArdle, and Sachs, 1987	< 6	0.06	Any	Yes	0.05	0.14	0.94	0.13	0.95
Gilson, O'Brien, Vera, et al., 1988	< 8	0.02	C-section for fetal distress	Yes	0.20	0.08	0.99	0.67	0.81
Gilson, O'Brien, Vera, et al., 1988	< 8	0.12	Fetal distress	Yes	0.20	0.15	0.89	0.27	0.81
Gilson, O'Brien, Vera, et al., 1988	< 8	0.10	Postmaturity syndrome	Yes	0.20	0.27	0.94	0.54	0.83

C-section = Cesarean section

NR = Not reported

Arabin, Snyjders, Mohnhaupt, et al. (1993) compared the predictive ability of a biophysical profile consisting of NST, amniotic fluid assessment, fetal tone, fetal movements, and fetal breathing to a novel fetal assessment score consisting of five components: FHR pattern, uterine artery resistance by Doppler ultrasound, carotid artery resistance index by Doppler ultrasound, fetal tone (movements) by ultrasound, and fetal reflexes (magnitude and speed of movements) by ultrasound. In receiver operating characteristic (ROC) analysis, the fetal assessment score provided better prediction of fetal distress and low Apgar score at 1 minute than did the biophysical profile (p < 0.001) but not better prediction of low umbilical artery pH. Qualitatively, the difference was greatest for prediction of fetal distress, with less difference noted for prediction of low Apgar scores and none for prediction of low pH. This suggests that the fetal prediction score is better at discriminating results that correlate directly with its component tests (such as fetal distress defined by abnormal fetal heart rate patterns) than at true physiological measures of fetal compromise. One possible explanation for this could be interpretation of intrapartum fetal monitoring based on prior knowledge of antepartum test results.

Hann, et al., reported the results of biophysical profile monitoring in 131 women at 41 completed weeks gestation (Hann, McArdle, and Sachs, 1987). Positive predictive values for "poor neonatal outcome" (neonatal distress requiring admission to the neonatal intensive care unit, endotracheal intubation, use of positive pressure ventilation for more than 6 hours, and/or persistent fetal circulation) for the composite biophysical profile at a threshold of < 6 was 14 percent; for individual components, positive predictive values were as follows: AFV, 17 percent; placental grading, 4 percent; fetal breathing movements, 5 percent; fetal tone/movements, 40 percent; and nonreactive NST, 14 percent. Negative predictive value for the composite biophysical profile was 94 percent; for individual components: AFV, 95 percent; placental grading, 91 percent; fetal breathing movements, 94 percent; fetal tone/movements, 95 percent; and reactive NST, 94 percent.

Gilson, O'Brien, Vera, et al. (1988) describe the association between twice weekly biophysical profile monitoring and low Apgar scores, fetal distress, and cesarean section for fetal distress among 178 women at greater than 42 weeks gestation. At the cut-point used (a score of 8), the test showed poor sensitivity across all outcomes, ranging from 0.08 to 0.27.

Doppler measurements of umbilical blood flow

Two studies reported data on the predictive value of Doppler measurments of umbilical artery blood flow (Battaglia, Larocca, Lanzani, et al., 1991). This was performed as a battery of tests including NST; amnioscopy; AFV; Doppler velocimetry of the uterine, umbilical, descending thoracic aorta, renal, and middle cerebral arteries; and a series of maternal blood measurements, including hPL, estriol, hematocrit, platelets, mean platelet volume, and uric acid. The criteria for decisionmaking about induction and delivery were not described. Doppler velocimetry was strongly associated with adverse outcomes, including "poor condition" (both 1- and 5-minute Apgar scores < 7 or infant admitted to NICU for asphyxia and/or meconium aspiration syndrome), oligohydramnios (largest pocket < 2 cm), meconium staining, and cesarean sections for fetal distress. Of note, 4 of 16 of these infants had birthweights greater than 4,000 grams; it is unclear to what extent these infants, who presumably had normal uteroplacental function, affected the results.

Farmakides, et al., reported on 140 high-risk pregnancies (33 percent were postdate) that were followed with NST and Doppler velocimetry (Farmakides, Schulman, Winter, et al., 1988). "Most" of the cases of fetal distress and cesareans for fetal distress came from the postdate subgroup. Nonreactive NST was significantly more sensitive at predicting cesarean section for fetal distress than Doppler. Since management decisions were based on NST results, this again raises the possibility of biased decisionmaking based on prior knowledge of antepartum test results.

Table 15: Doppler examination of umbilical artery flow (more...)

Table 15: Doppler examination of umbilical artery flow test performance characteristics

Study	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Any	Yes	0.01	1.00	0.72	0.04	1.00
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Cesarean section	Yes	0.29	0.58	0.83	0.58	0.83
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Cesarean section for fetal distress	Yes	0.12	0.80	0.78	0.33	0.97
Farmakides, Schulman, Winter, et al., 1988	0.31	Fetal distress	Abnormal	0.43	0.27	0.65	0.36	0.54
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Meconium staining	Yes	0.29	0.92	0.97	0.92	0.97
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Nonstress test	Abnormal	0.16	1.00	0.84	0.54	1.00
Battaglia, Larocca, Lanzani, et al., 1991	0.29	Oligohydramnios	Yes	0.30	0.64	0.86	0.67	0.84

Summary of tests to evaluate risks to the fetus associated with uteroplacental insufficiency

There are no randomized trials comparing antepartum testing by any method to no testing in women with prolonged pregnancy only. Data from one relatively large retrospective cohort (Bochner, Williams, Castro, et al., 1988) suggest an increased risk of adverse outcomes to the fetus, although confounding cannot be eliminated as a possibility for this observed association. Evidence from large registries shows consistently elevated risks of antepartum stillbirth with increasing gestational age, even in health systems where testing is available (see the section on "Risk of Perinatal Mortality" in chapter 1). Given this elevated risk, it is highly unlikely that a randomized trial of testing versus no testing could be performed in the United States without, at the least, extreme difficulty with recruitment. The low absolute risk of stillbirth makes sample size requirements prohibitive as well. For example, the estimated perinatal mortality at 41 weeks in terms of deaths per 1,000 ongoing pregnancies is approximately 1.2. A randomized trial would need over 40,000 women in each arm to determine a two-fold difference in risk of stillbirth between two competing methods of antepartum surveillance.

Because of the numerous methodological issues involved in evaluating specific antepartum tests (see discussion below), we are unable to conclude that any test or combination of tests is clearly superior to another. Only one randomized trial directly compared a more complex test with a simpler test (Alfirevic and Walkinshaw, 1995); this trial showed that the more complex test resulted in more interventions with no difference in outcomes. As with most tests, there appear to be consistent tradeoffs between sensitivity and specificity-tests that are more sensitive are likely to be less specific. We did not identify published data on inter- or intraobserver variability of these tests in the specific context of monitoring prolonged pregnancy or on the medical and nonmedical costs associated with specific tests and testing regimens.

We did find that, qualitatively, specificity for most tests was considerably better than sensitivity, while negative predictive value also was considerably better than positive predictive value. This means that women with "normal" test results are highly unlikely to experience the adverse outcomes used to determine a true "positive" test result. The high specificities reported may reflect biases in study design-when outcomes are either directly related to test results (such as nonreassuring fetal heart rate tracings after abnormal antepartum NST) or likely to be influenced by knowledge about the test results (such as cesarean section for fetal distress), specificity is likely to be relatively high.

This pattern of high negative predictive value in the setting of relatively low sensitivities has interesting implications for future management strategies. By Bayes' Theorem, positive predictive value can be expressed as:

True Positives/(True Positives + False Positives), or [(Prevalence)*(Sensitivity)] /{[(Prevalence)*(Sensitivity)] + [(1-Prevalence)*(1-Specificity)]}, while negative predictive value is expressed as:

True Negatives/(True Negative + False Negatives), or [(1-Prevalence)*(Specificity)] /{[(1-Prevalence)*(Specificity)] + [(Prevalence)*(1-Sensitivity)]}.

In practice, this means that increasing test sensitivity results in a higher negative predictive value, since the false negative rate decreases. Increasing test specificity results in a higher positive predictive value, since false positives decrease. Given the consistent pattern observed for all of the reviewed antepartum tests that specificity is higher than sensitivity, one would expect that positive predictive value would be higher than negative predictive value. The fact that the pattern is consistently the opposite suggests that it is the relatively low prior probability of adverse outcomes, the "prevalence" in the equations above, that drives the predictive values.

If this is the case, then the following points need to be considered:

• The main purpose of antepartum testing is primarily to avoid unexplained stillbirths and secondarily to avoid perinatal morbidity. In order to accomplish these things, tests with high negative predictive values are needed. One way to achieve this would be to improve the sensitivity of currently used antepartum testing technologies. Since it is unlikely that sensitivity can be increased without a subsequent decrease in specificity, this means that the positive predictive value of these tests will decrease further.

• If, as the reviewed studies suggest, the probability of adverse outcomes is currently what determines predictive values, then this means that the positive predictive value of antepartum testing will improve and the negative predictive value decline as gestational age increases, since the risk of stillbirth and other adverse events increases with gestational age. This proposition is dependent on the assumptions that (1) sensitivity and specificity are independent of gestational age, and (2) the outcomes reported in these studies are reasonable surrogates for stillbirth risk. This proposition is consistent with the data reported by Bochner, Williams, Castro, et al. (1988), according to which the positive predictive value for all adverse outcomes was better when testing began at 42 weeks (21.1 percent vs. 11.9 percent when testing began at 41 weeks), but the negative predictive value was worse (98.5 percent at 42 weeks vs. 99.1 percent at 41 weeks).

• Assuming that induction of labor does not carry increased perinatal risks compared with spontaneous labor, planned induction of labor at a given gestational age will always result in fewer expected adverse perinatal outcomes compared with testing strategies, since the negative predictive value of the tests will continue to decline as gestational age advances. At earlier gestational ages, where the risk is very low, the number of patients required to demonstrate this would be quite large.

These implications will be discussed further in the context of the trials of induction versus testing (Question 2).

Assessment of Risks to the Fetus and Mother Associated with Fetal Macrosomia

Because both mother and infant are at risk of injury secondary to macrosomia, various methods for estimating fetal weight have been evaluated. Macrosomia is usually defined as a newborn weight of greater than 4,000 grams or 4,500 grams; the clinical significance of birthweights between 4,000 and 4,500 grams is unclear, since risk of shoulder dystocia is greatest for infants over 4,500 grams (ACOG, 2000).

Ultrasound

Chauhan, et al. (Chauhan, Sullivan, Magann, et al., 1994) found that clinical estimation was more accurate than ultrasonographic estimation by the same clinician (see above). Ultrasound was slightly more sensitive at predicting birthweight greater than 4,000 grams (55 percent vs. 50 percent, based on 20 cases).

Chervenak, et al., compared 317 women followed for prolonged pregnancy with twice weekly NST and AFT with 100 control patients delivered between 38 and 40 weeks (Chervenak, Divon, Hirsch, et al., 1989). Fetal weights were also obtained, although it is unclear how often these measurements were performed. Overall incidence of birthweight greater than 4,000 grams was significantly higher in postdate patients (24 percent vs. 4 percent; p < 0.05), and cesarean section rates for arrest or protraction disorders were significantly higher when infants weighed more than 4,000 grams (22 percent vs. 10 percent; p < 0.01). Sensitivity of ultrasound for predicting birthweight greater than 4,000 grams was 61 percent, specificity 91 percent, positive predictive value 70 percent, and negative predictive value 87 percent. Morbidity associated with macrosomia was not reported. It is unclear to what extent clinicians managing the patients had access to the ultrasound reports. Since clinicians might have a lower threshold for diagnosing an arrest or protraction disorder in the setting of suspected macrosomia, this would result in a bias in favor of improved positive predictive value for ultrasound.

Gilby, et al., constructed ROC curves for the performance of two abdominal circumference cut-points (35 cm and 38 cm) for predicting macrosomia at two thresholds, 4,000 grams and 4,500 grams, from a series of 1,996 subjects who had ultrasounds within 7 days of delivery (Gilby, Williams, and Spellacy, 2000). At a cut-point of 35 cm, sensitivity for prediction of birthweight of 4,500 grams was 98.5 percent, specificity 64.6 percent, positive predictive value 9.1 percent, and negative predictive value 99.9 percent. At a cut-point of 38 cm, sensitivity was 53.6 percent, specificity 96.8 percent, positive predictive value 37.3 percent, and negative predictive value 98.3 percent. Morbidity associated with macrosomia was not reported. Whether these predictive values would be applicable in a different population is unclear.

O'Reilly-Green and Divon (1997) constructed ROC curves for ultrasonographic estimates of fetal weight, with an adjustment of 12.7 grams added to the estimated fetal weight (EFW) for each day elapsed between sonographic measurements and delivery. Areas under the ROC curve for prediction of birthweight greater than 4,000 grams were 0.85 and 0.93 to 0.95 for prediction of birthweight greater than 4,500 grams, indicating good discriminative ability. Relatively small relative increments in EFW had large impacts on sensitivity and specificity: for prediction of actual birthweight of greater than 4,000 grams, an EFW of 3,711 grams had a sensitivity of 85 percent and specificity of 72 percent, while an EFW of 4,000 grams had a sensitivity of 56 percent and a specificity of 91 percent. For prediction of birthweight greater than 4,500 grams, an EFW of 4,192 grams had sensitivity of 83 percent and specificity of 92 percent, while an EFW of 4,500 grams had a sensitivity of 22 percent and a specificity of 99 percent. Again, no correlation with outcomes associated with fetal macrosomia were reported.

Table 16: Accuracy of antenatal fetal weight estimation (more...)

Table 16: Accuracy of antenatal fetal weight estimation for predicting macrosomia

Study	Screening Test Threshold	Rate of Abnormal Tests	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Chauhan, Sullivan, Lutton, et al., 1995	> 4,000 gm	0.21	> 4,000 gm	0.26	0.61	0.92	0.73	0.87
O'Reilly-Green and Divon, 1997	> 3,710 gm	0.42	> 4,000 gm	0.24	0.85	0.72	0.49	0.94
Chervenak, Divon, Hirsch, et al., 1989	> 4,000 gm	0.22	> 4,000 gm	0.26	0.60	0.91	0.69	0.87
Pollack, Hauer-Pollack, and Divon, 1992	> 4,000 gm	0.20	> 4,000 gm	0.23	0.56	0.91	0.65	0.88
Jazayeri, Heffron, Phillips, et al., 1999	Abdominal circumference > 34 cm	0.48	> 4,000 gm	0.50	0.89	0.93	0.93	0.90
Gilby, Williams, and Spellacy, 2000	Abdominal circumference > 38 cm	0.05	> 4,500 gm	0.03	0.54	0.97	0.37	0.98
O'Reilly-Green and Divon, 1997	> 4500 gm	0.02	> 4,500 gm	0.04	0.22	0.99	0.44	0.97
O'Reilly-Green and Divon, 1997	> 4,191 gm	0.11	> 4,500 gm	0.04	0.83	0.92	0.30	0.99
Pollack, Hauer-Pollack, and Divon, 1992	> 4,500 gm	0.02	> 4,500 gm	0.04	0.14	0.99	0.33	0.96
Gilby, Williams, and Spellacy, 2000	Abdominal circumference >34 cm	0.38	> 4,500 gm	0.03	0.99	0.65	0.09	1.00

Test performance characteristics for studies reporting association between estimated fetal weight and macrosomia are shown in Table 16.

Assessment of the Likelihood of Successful Induction

Cervical examination (Bishop score)

Table 17: Components of the Bishop score

Table 17: Components of the Bishop score

Exam Findings	Score
	0	1	2	3
Dilatation (cm)	Closed	1-2	3-4	> 5
Effacement (%)	0-30	40-50	60-70	> 80
Station	−3	−2	−1 or 0	+1, +2
Consistency	Firm	Medium	Soft	-
Position of cervix	Posterior	Midposition	Anterior	-

The Bishop score was first reported in 1964 as a predictor of the likelihood of a successful induction (Bishop, 1964). The score is based on five components: cervical dilation, cervical effacement, cervical consistency, cervical position, and fetal station (Table 17).

In Bishop's original report (Bishop, 1964), induction was successful in 100 percent of cases (no denominator given) when the Bishop score was greater than 9. Data for lower scores were not given, and notably, all inductions were apparently in multiparous patients, since "[o]wing to the unpredictability of the duration of labor in the nullipara, even in the presence of apparently favorable circumstances, induction of labor brings little advantage for either obstetrician or patient." There was a statistically significant negative correlation between score and interval from examination to spontaneous delivery, but confidence intervals were quite wide (quantitative data were not provided, only a graphic representation).

Table 18: Bishop score test performance characteristics (more...)

Table 18: Bishop score test performance characteristics

Study	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Mouw, Egberts, Kragt, et al., 1998	NR	Birth date	< 4 days	0.5	0.67	0.77	NR	NR
Witter and Weitz, 1989	0.41	Cesarean section	Yes	0.34	0.77	0.78	0.65	0.87
Witter and Weitz, 1989	0.24	Cesarean section	Yes	0.34	0.46	0.88	0.67	0.76
Witter and Weitz, 1989	0.54	Cesarean section	Yes	0.34	0.73	0.56	0.46	0.80

NR = Not reported

Three studies provided limited data on the predictive value of Bishop scores (Harris, Huddleston, Sutliff, et al., 1983; Mouw, Egberts, Kragt, et al., 1998; Witter and Weitz, 1989). Harris, et al., reported that dilatation, effacement, and station were more predictive of interval between examination and spontaneous delivery in prolonged pregnancy than consistency and position (Harris, Huddleston, Sutliff, et al., 1983). Witter and Weitz (1989) found that Bishop scores at baseline in women induced at 42 weeks were statistically significantly lower in women who underwent cesarean delivery than in those with vaginal delivery, but that the absolute difference was small; significant overlap made the test a poor discriminator of successful induction (Table 18). Mouw, et al., reported that a Bishop score greater than 5 at 41 weeks had sensitivity 0.67 (95 percent CI, 0.48 to 0.82) and specificity 0.77 (95 percent CI, 0.54 to 0.92) for predicting birth within 3 days;however, only 74 percent of patients in this study had Bishop scores recorded (Mouw, Egberts, Kragt, et al., 1998).

The relatively poor discrimination of the Bishop score in predicting either labor or subsequent successful induction in prolonged pregnancy is magnified by the inherent unreliability of many of its component measures. Significant interobserver variability has been reported in measurement of cervical effacement (Goldberg, Newman, and Rust, 1997; Holcomb and Smeltzer, 1991). Furthermore, significant intra- and interobserver variability has been described for assessment of cervical dilatation (Phelps, Higby, Smyth, et al., 1995; Tuffnell, Bryce, Johnson, et al., 1989)

Fibronectin

Table 19: Fetal fibronectin test performance characteristics (more...)

Table 19: Fetal fibronectin test performance characteristics

Study	Screening Test Threshold	Rate of Abnormal Tests	Outcome	Outcome Threshold	Rate of Outcome Event	Sensitivity	Specificity	Positive Predictive Value	Negative Predictive Value
Tam, Tai, and Rogers, 1999	Positive	0.33	Vaginal Delivery	Yes	0.84	0.36	0.84	0.86	0.32
Mouw, Egberts, Kragt, et al., 1998	> 50 ng/ml	0.54	Birth date	< 4 days	0.50	0.71	0.64	0.67	0.69

Three studies were identified that evaluated the possible use of fetal fibronectin (fFN) obtained from cervicovaginal secretions, a sensitive marker for impending labor, in the management of prolonged pregnancies (Table 19). Tam, et al., measured fetal fibronectin in 58 women at term or beyond, scheduled for induction with PGE₂ suppositories (Tam, Tai, and Rogers, 1999). Thirty women were negative and 28 positive for fibronectin prior to the placement of the suppositories. There was a trend towards a higher gestational age in fibronectin-positive patients (median 294 days, range 280-294, compared with a median of 281 days, range 272-294, in negative patients). Median interval from induction to delivery was significantly lower in fibronectin-positive patients (760 minutes vs. 1,285 minutes). Fibronectin positivity was a reasonable predictor of vaginal delivery (sensitivity 36 percent; specificity 79 percent; positive predictive value 84 percent; negative predictive value 28 percent). Results in this study were not stratified by gestational age or by indication for induction.

Mouw, et al., measured fetal fibronectin at 41 weeks (Mouw, Egberts, Kragt, et al., 1998). A positive fFN test (> 50 ng/ml) had sensitivity of 0.71 (95 percent CI, 0.58 to 0.86) and specificity of 0.64 (95 percent CI, 0.48 to 0.78) for predicting birth within 3 days. The change from negative to positive fFN values often occurred between 1 and 4 days before birth in women with a spontaneous onset of labor. The mean interval between positive test and birth was 2.5 ± 2.5 days (range, 0-11).

Imai and colleagues measured vaginal fFN and a panel of cytokines (interleukin 1-beta, interleukin-6, interleukin-8, and tumor necrosis factor alpha) weekly in 122 women from 36 through 42 weeks (Imai, Tani, Saito, et al., 2001). Vaginal fFN was inversely correlated with sampling to delivery interval (r = −0.40). At a threshold of > 50 ng/ml, fFN had a sensitivity of 90 percent, a specificity of 50 percent, a positive predictive value of 75 percent, and a negative predictive value of 75 percent for predicting delivery within 7 days. Interleukin 1-beta was the only cytokine with reasonable performance, but it was less able to discriminate than fFN (sensitivity 55 percent, specificity 76 percent). Results were not stratified by parity or gestational age.

Study Design

• Choice of appropriate outcome measures: Many of the most important outcome measures, especially stillbirth, are so rare that studies using these outcomes are almost impossible to perform. Surrogate markers therefore are not inappropriate, but their clinical relevance is not always clear. For example, although meconium aspiration is a significant adverse outcome with potential for long-term negative sequelae, the presence of meconium-stained amniotic fluid alone is not. Intrapartum abnormal fetal heart rate tracings themselves are subject to significant observer variability (Ayres-de-Campos, Bernardes, Costa-Pereira, et al., 1999; Bernardes, Costa-Pereira, Ayres-de-Campos, et al., 1997; Donker, van Geijn, and Hasman, 1993; Lidegaard, Bottcher, and Weber, 1992), and interpretation may be influenced by prior knowledge of antepartum test results, making fetal heart rate patterns, or cesarean section decisions based on these patterns, less than ideal as surrogate markers of fetal compromise.

• Bias: Many of the studies reviewed either did not state whether clinicians managing patients were aware of test results or definitely stated that these results were available. Since knowledge of these results could affect both interpretation of outcomes (as discussed above) or thresholds for decisionmaking (e.g., greater reluctance to use oxytocin to augment labor if prior antepartum testing was abnormal, or a lower cesarean section threshold for arrest of dilatation or descent if macrosomia were suspected), the ability of tests to predict these outcomes could be falsely elevated.

• Resource use: Data on the medical and nonmedical costs of any of the tests reviewed are lacking.

Statistical Issues

• Inappropriate summary measures and tests: Many studies used means or t-tests for variables such as Bishop scores, Apgar scores, or parity, where values other than integers are meaningless.

• Sample size: Few studies discussed sample size issues.

• Failure to account for variability: No study attempted to account for the effects of observer variation on the precision of estimates. For tests where quantitative values are used to establish a threshold for normal and abnormal, this variability will have implications for the precision of sensitivity and specificity.

Trials Identified

The included trials were published between 1983 and 1997. The number of subjects in each trial was fairly small, except for the Canadian trial (Hannah, Hannah, Hellmann, et al., 1992). The overall median number of subjects was 200, ranging from 22 (Martin, Sessums, Howard, et al., 1989) to 3,418 (Hannah, Hannah, Hellmann, et al., 1992).

Benefits

Effects on perinatal mortality

Table 20: Perinatal mortality (excluding deaths due (more...)

Table 20: Perinatal mortality (excluding deaths due to congenital abnormalities)

Study	Induction	Monitoring	Cause(s) of Death
Augensen, Bergsjø, Eikeland, et al., 1987	0/214	0/195	-
Bergsjø, Huang, Yu, et al., 1989	0/94	1/94	Pneumonia
Cardozo, Fysh, and Pearce, 1986	0/195	1/156	Stillbirth
Dyson, Miller, and Armstrong, 1987	0/152	1/150	Infant was delivered by emergency C-section 14 minutes after onset of bradycardia, with venous cord pH = 7.04; subsequently died of complications of meconium aspiration and persistent fetal circulation
Egarter, Kofler, Fitz, et al., 1989	0/180	1/165	Stillbirth (died because of a "cord complication")
El-Torkey and Grant, 1992	0/33	0/32	-
Hannah, Hannah, Hellmann, et al., 1992	0/1701	2/1706	Both stillbirths
Heden, Ingemarsson, Ahlstrom, et al., 1991	NR	NR	-
Herabutya, Prasertsawat, Tongyai, et al., 1992	0/57	0/51	-
Katz, Yemini, Lancet, et al., 1983	0/78	1/78	Emergency C-section performed; baby died 24 hours after birth with autopsy showing asphyxia.
Martin, Sessums, Howard, et al., 1989	0/12	0/10	-
National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units, 1994	0/265	0/175	-
Ohel, Rahav, Rothbart, et al., 1996	NR	NR	-
Witter and Weitz, 1987	NR	NR	-

NR = not reported

The included studies suggest that induction results in fewer perinatal deaths than does expectant management. Table 20 summarizes perinatal deaths not due to congenital abnormalities in the two management groups. There were a total of seven deaths in the monitoring group compared with no deaths in the induction group.

A meta-analysis performed as part of a recent Cochrane review (Crowley, 2000) showed that this reduction in perinatal mortality with induction is significant only at 41 weeks or later (summary odds ratio [OR], 0.13; 95 percent confidence interval [CI], 0.01 to 2.07 before 41 weeks vs. summary OR, 0.23; 95 percent CI, 0.06 to 0.90 at 41 weeks or later).

Risks

Cesearean section

Table 21: Overall cesarean section rates†

Table 21: Overall cesarean section rates†

Study	Induction Group	Monitoring Group
Augensen, Bergsjø, Eikeland, et al., 1987	14/214 (6.5%)	15/195 (7.7%)
Bergsjø, Huang, Yu, et al., 1989	27/94 (28.7%)*	39/94 (41.5%)*
Cardozo, Fysh, and Pearce, 1986	25/195 (13%)	18/207 (9%)
Dyson, Miller, and Armstrong, 1987	22/152 (14.5%)*	41/150 (27.3%)*
Egarter, Kofler, Fitz, et al., 1989	2/180 (1.1%)	3/165 (1.8%)
El-Torkey and Grant, 1992	5/33 (15%)	4/32 (12.5%)
Hannah, Hannah, Hellmann, et al., 1992	360/1701 (21.2%)*	418/1706 (24.5%)*
Heden, Ingemarsson, Ahlstrom, et al., 1991	10/109 (9.2%)	9/127 (7.0%)
Herabutya, Prasertsawat, Tongyai, et al., 1992	27/57 (47.4%)	24/51 (47.1%)
Katz, Yemini, Lancet, et al., 1983	16/78 (20.5%)*	7/78 (8.8%)*
Martin, Sessums, Howard, et al., 1989	2/12 (17 %)*	1/10 (10 %)*
National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units, 1994	55/265 (20.8%)	32/175 (18.3%)
Ohel, Rahav, Rothbart, et al., 1996	4/70 (5.7%)	6/104 (5.8%)
Witter and Weitz, 1987	30/103 (29.1%)	27/97 (27.8%)

†

Rates given represent overall cesarean section rates except in the case of Cardozo, Fysh, and Pearce (1986), which reported only emergency cesarean section rates.

* Statistically significant difference

Of the 15 included trials, two found a statistically increased risk of overall cesarean section with induction, while three trials found a statistically increased risk of overall cesarean section with expectant monitoring (Table 21).

Table 22: Summary odds ratios for cesarean section (more...)

Table 22: Summary odds ratios for cesarean section in randomized trials of elective induction versus expectant management†

Group	Odds Ratio	95% Confidence Interval
By gestational age:
< 41 weeks	0.6	0.35 to 1.03
> 41 weeks	0.87	0.77 to 0.99
By parity:
Primigravid	0.75	0.64 to 0.88
Multigravid
Bishop Score < 6	1.02	0.75 to 1.38
By overall cesarean section rate:
< 10%	0.89	0.58 to 1.39
> 10%	0.87	0.77 to 1.00
By class of induction agent:
Oxytocin	0.85	0.74 to 0.98
Prostaglandins	0.84	0.65 to 1.08

†

Data from Crowley (2000); odds ratios < 1 indicate that the cesarean section rate was lower in the elective induction group.

Meta-analysis and subgroup analyses performed as part of a recent Cochrane review (Crowley, 2000) found no significant differences in cesarean delivery rates in any group or subgroup (Table 22). If anything, cesarean rates tend to be slightly lower in the elective induction groups.

Hannah, et al., published an interesting reanalysis of the Canadian study in 1996 (Hannah, Huh, Hewson, et al., 1996). In this new analysis, women who were randomized to induction or expectant management were stratified based on whether labor was ultimately induced or spontaneous. In the induction arm, 772/1,149 women (67.7 percent) were induced, while 377/1,149 (33.3 percent) went into spontaneous labor prior to scheduled induction. In the expectant management group, 405/1,128 (35.9 percent) were induced for various indications, while 723/1,128 (64.1 percent) went into spontaneous labor. There were no significant differences in cesarean section rates between women randomized to induction who were induced (29.5 percent), women randomized to induction who went into spontaneous labor (25.7 percent), and women who were managed expectantly who went into spontaneous labor (25.7 percent). However, the cesarean section rate was significantly increased in women randomized to expectant management who were induced (42.0 percent). These women were significantly more likely to be nulliparous, to have a closed cervix at the onset of labor, and to have a longer interval from induction to delivery. When compared with the expectantly managed women in spontaneous labor, they had significantly higher cesarean section rates for fetal distress or dystocia; such differences were not seen when the two subgroups in the induction arm were compared.

These differences are consistent with several findings discussed earlier in this report:

• Women whose onset of labor is considerably later than average may represent a distinct subgroup with different physiological characteristics of the uterus and cervix. This is consistent with the higher proportion of women with closed cervices and may also explain the higher rates of cesarean section for dystocia. This also may be related to parity. Presumably, women are included in this group who reach a predefined date for induction without going into spontaneous labor and with normal antepartum testing.

• Provider knowledge of antepartum testing results may affect thresholds for cesarean delivery. It seems likely that providers caring for women whose inductions were indicated because of abnormal antepartum tests would be less tolerant of intrapartum fetal heart rate abnormalities or less likely to tolerate labor progress that was slower than average. This would explain some of the differential rates by indication.

• As Crowley (2000) points out, women induced in the expectant management arm were less likely to receive prostaglandins. This would be a bias in favor of induction. The reanalysis by Hannah and colleagues (Hannah, Huh, Hewson, et al., 1996) models this based on assumptions about prostaglandin efficacy, and finds that, at worst, there would be no difference in cesarean section rates between groups. In addition, our review of the literature on induction agents (discussed under Question 3) suggests that the effectiveness of prostaglandins in terms of expediting delivery may be proportional to risk of fetal heart rate abnormalities in labor. If this is the case, then any decrease in cesarean section rates for failed induction or dystocia might well be accompanied by an increase in cesarean sections for fetal distress.

In summary, the randomized trial literature consistently shows that elective induction does not result in increased cesarean section rates compared with management strategies based on antepartum testing. If anything, cesarean section rates are slightly lower in women who are electively induced.

Costs and Resource Use

Indirect measures of resource use

Several studies that did not report direct costs did report outcomes that are indirect measures of resource use, such as overall length of maternal or infant stay in the hospital. The extent to which these results are generalizable is limited, since length of stay varies internationally and has changed dramatically in the United States over recent years. Moreover, overall length of stay may not be entirely related to overall resource use (Tai-Seale, Rodwin, and Wedig, 1999). For women delivering in a hospital, the majority of resource use occurs during the time from admission to delivery, with a sharp decrease after delivery and even further decreases after the first 24 hours. Thus, even if the mean length of stay is equivalent between two groups, the resource use may vary widely depending on what proportion of the time was spent in the delivery suite. In addition, studies that report only hospital use and not outpatient use of resources (for antepartum testing, other office visits, etc.) will not reflect the overall medical costs of a particular strategy. Finally, none of the included studies addressed the nonmedical costs -- such as transportation, time lost from work, child care for women with other children, and so on -- associated with various strategies for managing prolonged pregnancy.

Table 23: Mean maternal length of stay, induction versus (more...)

Table 23: Mean maternal length of stay, induction versus expectant management

Study	Mean Maternal Length of Stay (in Days)		P-value
	Induction	Expectant Management
Augensen, Bergsjø, Eikeland, et al., 1987	7.05	6.69	< 0.02
Martin, Sessums, Howard, et al., 1989	3.4	2.6	NS
Witter and Weitz, 1987	4.74	4.06	< 0.02
Bergsjø, Huang, Yu, et al., 1989	7.9	8.1	Not reported
Dyson, Miller, and Armstrong, 1987	3.2	3.5	< 0.04
Hannah, Hannah, Hellman, et al., 1992	3.9	4.0	Not reported

NS = Not significant

Table 23 shows reported mean maternal lengths of stay for the six trials where this was reported. There are no obvious trends. Because reporting of the proportion of time spent in labor versus postpartum was minimal, no additional inferences about relative resource use can be drawn.

Only one study (Dyson, Miller, and Armstrong, 1987) reported data on mean neonatal length of stay, with no significant differences between the induction and expectant management groups (3.0 days vs. 3.3 days, respectively).

Table 24: Perinatal outcomes, induction versus monitoring (more...)

Table 24: Perinatal outcomes, induction versus monitoring

Study	Gesta-tional age*	N	Induction Method	Monitoring Method	Perinatal Outcomes
					Induction	Monitoring	P-value
Ohel, Rahav, Rothbart, et al., 1996; Ohel, Rahav, Rothbart, et al., 1996	40 weeks	200	PGE2	Fetal monitoring: 2x/week	Perinatal death: NR Apgar (mean @ 5 min) 9.5 NICU admission: NR	Perinatal death: NR Apgar (mean@ 5 min) 9.4 NICU admission: NR	NS
Egarter, Kofler, Fitz, et al., 1989	40 weeks	356	PGE2, repeated if necessary	Nonstress test: q 2-3 days	Perinatal death: 0 Apgar - low scores were not different between gps NICU admission: NR	Perinatal death: 1 Apgar (see induction) NICU admission: NR	NR
El-Torkey and Grant, 1992	41 weeks	65	Membrane sweeping	No vaginal examination	Perinatal death: 0 Apgar < 6 @ 5 min: 3% NICU admission: NR	Perinatal death:) Apgar < 6@5 min: 3% NICU admission: NR	NR 0.98
Dyson, Miller, and Armstrong, 1987	41 weeks	302	PGE2 as outpatient, repeated if necessary, then oxytocin if needed	Nonstress test: 2x/week, Amniotic fluid index: once between 41 and 42 weeks, twice weekly after 42 weeks	Perinatal death: 0 Apgar < 7 @ 5 min: 11.2% NICU admission: NR	Perinatal death: 1 Apgar < 7 @ 5 min: 21.3% NICU admission: NR	NS <0.02
National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units, 1994	41 weeks	440	1) PGE2 only 2) PGE2 plus oxytocin	Nonstress test: 2x/week AFI: 2x/week	Perinatal death: 0 Apgar < 4 @ 5 min: 0 NICU admission: NR	Perinatal death: 0 Apgar<4@5 min: <1% NICU admission: NR	NR NR
Hannah, Hannah, Hellmann, et al., 1992	41 weeks	3418	PGE2, oxytocin	Kick counts once a day Nonstress test: 3x/week Amniotic fluid index: 2-3x/week	Perinatal death: 0 Apgar < 7 @ 5 min: 1.1% NICU admission:4%	Perinatal death: 2 Apgar<7@5min: 1.2% NICU admission: 1%	NR NS NS
Cardozo, Fysh, and Pearce, 1986	41-5/7 weeks	402	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Perinatal death: 0 Apgar < 5@ 5 min: 1% NICU admission: 3%	Perinatal death: 1 Apgar < 7@5 min: 2% NICU admission:1.5%	NR NS NS
Pearce and Cardozo, 1988	41-5/7 weeks	281	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Perinatal death: 0 Apgar < 5 @ 5 min: 1% NICU admission: 4%	Perinatal death: 1 Apgar < 5@ 5 min:1% NICU admission: 1%	NR NS NS
Augensen, Bergsjø, Eikeland, et al., 1987	41-5/7 weeks	409	Oxytocin, amniotomy	Nonstress test: immediate NST, repeated 3-4 days if undelivered	Perinatal death: 0 Apgar scores were "evenly distributed" NICU admission: 5.6%	Perinatal death: 0 Apgar (see induction) NICU admission: 7.7%	NR NR NR
Martin, Sessums, Howard, et al., 1989	42 weeks	22	Laminaria, oxytocin	NST/CST weekly Amniotic fluid index: weekly	Perinatal death: 0 Apgar (mean) 9.75 NICU admission: NR	Perinatal death: 0 Apgar (mean): 9.7 NICU admission: NR	NR NS
Herabutya, Prasertsawat, Tongyai, et al., 1992	42 weeks	108	PGE2, amniotomy, oxytocin	NST: weekly until 43 weeks, then 2x/week	Perinatal death: 0 Apgar < 7 @ 5 min: 1.8% NICU admission:1.8%	Perinatal death: 0 Apgar < 7 @ 5 min: 7.8% NICU admission: 7.8%	NR 0.19 0.19
Katz, Yemini, Lancet, et al., 1983	42 weeks	156	Amniotomy, oxytocin	Fetal monitoring: twice daily fetal movement counts Amnioscopy and OCT every 3 days	Perinatal death: 1.3% Apgar < 7 @ 5 min: 3.8% NICU admission: NR	Perinatal death: 1.3% Apgar< 7@ 5min: 1.3% NICU admission: NR	NS NR
Bergsjø, Huang, Yu, et al., 1989	42 weeks	188	Membrane stripping and oxytocin	Monitored group was admitted to hospital for one week w/close daily clinical surveillance "fetal movement test, atropine test, ultrasound and urinary estriol excretion tests were also employed"	Perinatal death: 0 Apgar no quantitative data "distributions were almost equal between the groups" NICU admission: NR	Perinatal death: 1 Apgar (see induction) NICU admission: NR
Witter and Weitz, 1987	42 weeks	200	Oxytocin, amniotomy	Fetal movement counts 3x/day OCT if decreased FM Urinary estriol once (41-42 wks), twice during 42-43 wks, and 3x if >43 wks	Perinatal death: NR Apgar < 7 @ 5 min: 0 NICU admission: NR	Perinatal death: NR Apgar < 7 @ 5 min: 2.08% NICU admission: NR	NS
Hedén, Ingemarsson, Ahlström, et al., 1991	42 weeks	238	Amniotomy, oxytocin	Nonstress test every other day Amniotic fluid index: once weekly Oxytocin challenge test: if nonreactive nonstress test	Perinatal death: NR Apgar < 7 @ 5 min: 2.8% NICU admission: 9.2%	Perinatal death: NR Apgar <7@5 min: 0.8% NICU admission: 6.2%	NS NS

*At randomization

NICU = Neonatal intensive care unit; NR = Not reported; NS = Not significant

OCT = Oxytocin challenge test; FM = Fetal movement

Table 25: Maternal outcomes, induction vs. monitoring (more...)

Table 25: Maternal outcomes, induction vs. monitoring

Study	Gestational age*	N	Induction Method	Monitoring Method	Maternal Outcomes
					Induction	Monitoring	P-value
Ohel, Rahav, Rothbart, et al., 1996	40 weeks	200	PGE2	Fetal monitoring: 2x/week	Cesarean section: 5.7% Operative vaginal delivery: NR	Cesarean section: 5.8% Operative vaginal delivery: NR	NS
Egarter, Kofler, Fitz, et al., 1989	40 weeks	356	PGE2, repeated if necessary	Nonstress test: q 2-3 days	Cesarean section: 1.1% Operative vaginal delivery: 2%	Cesarean section: 1.8% Operative vaginal delivery: 1.8%	NR NR
El-Torkey and Grant, 1992	41 weeks	65	Membrane sweeping	No vaginal examination	Cesarean section: 15% Operative vaginal delivery: 6%	Cesarean section: 12.5% Operative vaginal delivery: 9%	0.76 0.62
Dyson, Miller, and Armstrong, 1987	41 weeks	302	PGE2 as outpatient, repeated if necessary, then oxytocin if needed	Nonstress test: 2x/week, Amniotic fluid index: once between 41 and 42 weeks, twice weekly after 42 weeks	Cesarean section: 14.5% Operative vaginal delivery: NR	Cesarean section: 27.3% Operative vaginal delivery: NR	< 0.01
National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units, 1994	41 weeks	440	1) PGE2 only 2) PGE2 plus oxytocin	Nonstress test: 2x/week AFI: 2x/week	Cesarean section: 20.8% Operative vaginal delivery: NR	Cesarean section: 18.3% Operative vaginal delivery: NR	NS
Hannah, Hannah, Hellmann, et al., 1992	41 weeks	3418	PGE2, oxytocin	Kick counts: once a day Nonstress test: 3x/week Amniotic fluid index: 2-3x/week	Cesarean section: 21.2% Operative vaginal delivery: 35.3%	Cesarean section: 24.5% Operative vaginal delivery: 34.9%	= 0.03 NR
Pearce and Cardozo, 1988	41-5/7 weeks	281	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Cesarean section: 3% Operative vaginal delivery: 26%	Cesarean section: 1% Operative vaginal delivery: 28%	NS NS
Cardozo, Fysh, and Pearce, 1986	41-5/7 weeks	402	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Cesarean section: 13% Operative vaginal delivery: 20% Only reported Emergency C/S	Cesarean section: 9% Operative vaginal delivery: 26%	NS NS
Augensen, Bergsjø, Eikeland, et al., 1987	41-5/7 weeks	409	Oxytocin, amniotomy	Nonstress test: immediate NST, repeated 3-4 days if undelivered	Cesarean section: 6.5% Operative vaginal delivery: 17.2%	Cesarean section: 7.7% Operative vaginal delivery: 20.5%	NR NR
Martin, Sessums, Howard, et al., 1989	42 weeks	22	Laminaria, oxytocin	NST/CST weekly Amniotic fluid index: weekly	Cesarean section: 17% Operative vaginal delivery: 25%	Cesarean section: 10% Operative vaginal delivery: 25%	NS NS
Herabutya, Prasertsawat, Tongyai, et al., 1992	42 weeks	108	PGE2, amniotomy, oxytocin	NST: weekly until 43 weeks, then 2x/week	Cesarean section: 47.4% Operative vaginal delivery: 19.3%	Cesarean section: 47.1% Operative vaginal delivery: 17.6%	0.87 0.98
Katz, Yemini, Lancet, et al., 1983	42 weeks	156	Amniotomy, oxytocin	Fetal monitoring: twice daily fetal mvmt counts Amnioscopy and OCT every 3 days	Cesarean section: 20.5% Operative vaginal delivery: NR	Cesarean section: 8.8% Operative vaginal delivery: NR	< 0.05
Bergsjø, Huang, Yu, et al., 1989	42 weeks	188	Membrane stripping and oxytocin	Monitored gp was admitted to hospital for one week w/close daily clinical surveillance "fetal movement test, atropine test, ultrasound and urinary estriol excretion tests were also employed"	Cesarean section: 28.7% Operative vaginal delivery: 22.4%	Cesarean section: 41.5% Operative vaginal delivery: 26.6% *only p value reported compared operative deliveries (c-section. Forceps, and vacuum)	< 0.05
Witter and Weitz, 1987	42 weeks	200	Oxytocin, amniotomy	Fetal movement counts 3x/day OCT if decreased FM Urinary estriol once (41-42 wks), twice during 42-43 wks, and 3x if >43 wks	Cesarean section: 29.13% Operative vaginal delivery: NR	Cesarean section: 27.83% Operative vaginal delivery: NR	NS
Hedén, Ingemarsson, Ahlström, et al., 1991	42 weeks	238	Amniotomy, oxytocin	Nonstress test every other day Amniotic fluid index: once weekly Oxytocin challenge test: if nonreactive nonstress test	Cesarean section: 9.2% Operative vaginal delivery: 2.8%	Cesarean section: 7.0% Operative vaginal delivery: 15.5%	NS < 0.01

*At randomization

NR = Not reported; NS = Not significant

OCT = Oxytocin challenge test; FM = Fetal movement

Table 26: Resource use, induction versus monitoring (more...)

Table 26: Resource use, induction versus monitoring

Study	Gesta-tional age*	N	Induction Method	Monitoring Method	Resource Utilization
					Induction	Monitoring	P-value
Ohel, Rahav, Rothbart, et al., 1996	40 weeks	200	PGE2	Fetal monitoring: 2x/week	Total costs: NR Length of stay: NR Time to delivery (days to delivery from entrance into trial = 1.6 ± 1.9	Total costs: NR Length of stay: NR Time to delivery (see induction) = 5.2 ± 3.7	NR < 0.001
Egarter, Kofler, Fitz, et al., 1989	40 weeks	356	PGE2, repeated if necessary	Nonstress test: every 2-3 days	Total costs: NR Length of stay: NR Time to delivery (onset of contractions to delivery = 7.1 hours	Total costs: NR Length of stay: NR Time to delivery = 10.6 hrs p-value not reported, but authors report "statistically significant"	See text
El-Torkey and Grant, 1992	41 weeks	65	Membrane sweeping	No vaginal examination	Total costs: NR Length of stay: NR Time to delivery: NR	Total costs: NR Length of stay: NR Time to delivery: NR
Dyson, Miller, and Armstrong, 1987	41 weeks	302	PGE2 as outpatient, repeated if necessary, then oxytocin if needed	Nonstress test: 2x/week, Amniotic fluid index: once between 41 and 42 weeks, twice weekly after 42 weeks	Total costs: NR Length of stay (maternal hospital stay/mean = 3.2 days Time to delivery: NR	Total costs: NR Length of stay: 3.5 days Time to delivery: NR	< 0.04 < 0.001
National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units, 1994	41	440	1) PGE2 only 2) PGE2 plus oxytocin	Nonstress test: 2x/week AFI: 2x/week	Total costs: NR Length of stay: NR Time to delivery: 36 hours and 35 hours (2 induction gps)	Total costs: NR Length of stay: NR Time to delivery: 85 hours	< 0.001
Hannah, Hannah, Hellmann, et al., 1992	41 weeks	3418	PGE2, oxytocin	Kick counts once a day Nonstress test: 3x/week Amniotic fluid index: 2-3x/week	Total costs: $ 2502 Length of stay: 3.9 days Time to delivery: "women in the induction group were less likely not to have delivered their babies seven or more days after randomization" 5.1% vs 26% p<0.001	Total costs: $2684 Length of stay: 4.0 days Time to delivery: see induction
Pearce and Cardozo, 1988	41-5/7 weeks	281	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Total costs: NR Length of stay: NR Time to delivery: NR	Total costs: NR Length of stay: NR Time to delivery: NR
Cardozo, Fysh, and Pearce, 1986	41-5/7 weeks	402	PGE2 and oxytocin	Kick counts daily Ultrasound: AFI once, US also done to determine ratio of head circumference to abdominal circumference Fetal monitoring: NST every other day	Total costs: NR Length of stay: NR Time to delivery: NR	Total costs: NR Length of stay: NR Time to delivery: NR
Augensen, Bergsjø, Eikeland, et al., 1987	41-5/7 weeks	409	Oxytocin, amniotomy	Nonstress test: immediate NST, repeated 3-4 days if undelivered	Total costs: NR Length of stay: 7.05 days Time to delivery: 1.4 days	Total costs: NR Length of stay: 6.69 days Time to delivery: 4.0 days	0.02 NR
Martin, Sessums, Howard, et al., 1989	42 weeks	22	Laminaria, oxytocin	NST/CST weekly Amniotic fluid index: weekly	Total costs: NR Length of stay: 3.41 days Time to delivery (length of labor): 6.3 hours	Total costs: NR Length of stay: 2.6 days Time to delivery: 8.3 hours	NS NS
Herabutya, Prasertsawat, Tongyai, et al., 1992	42 weeks	108	PGE2, amniotomy, oxytocin	NST: weekly until 43 weeks, then 2x/week	Total costs: NR Length of stay: NR Time to delivery: NR	Total costs: NR Length of stay: NR Time to delivery: NR
Katz, Yemini, Lancet, et al., 1983	42 weeks	156	Amniotomy, oxytocin	Fetal monitoring: twice daily fetal mvmt counts Amnioscopy and OCT every 3 days	Total costs: NR Length of stay: NR Time to delivery (duration of labor): 9.4 hrs	Total costs: NR Length of stay: NR Time to delivery: 6.7 hrs	< 0.01
Bergsjø, Huang, Yu, et al., 1989	42 weeks	188	Membrane stripping and oxytocin	Monitored gp was admitted to hospital for one week w/close daily clinical surveillance "fetal movement test, atropine test, ultrasound and urinary estriol excretion tests were also employed"	Total costs: NR Length of stay: 7.9 days Time to delivery: -reported only in vaginal deliveries (no c/s), duration of labor from start of painful contractions of 10 min interval to birth was = 12 hrs 10 mins	Total costs: NR Length of stay: 8.1 days Time to delivery (see induction): 10 hrs 38 min	NR
Witter and Weitz, 1987	42 weeks	200	Oxytocin, amniotomy	Fetal movement counts 3x/day OCT if decreased FM Urinary estriol once (41-42 wks), twice during 42-43 wks, and 3x if >43 wks	Total costs: NR Length of stay: 4.06 Time to delivery: NR	Total costs: NR Length of stay: 4.74 Time to delivery: NR	< 0.05
Hedén, Ingemarsson, Ahlström, et al., 1991	42 weeks	238	Amniotomy, oxytocin	Nonstress test every other day Amniotic fluid index: once weekly Oxytocin challenge test: if nonreactive nonstress test	Total costs: NR Length of stay: NR Time to delivery: NR	Total costs: NR Length of stay: NR Time to delivery: NR

* At randomization

NR = Not reported

OCT = Oxytocin challenge test; FM = Fetal movement

Tables 24, 25, and 26 summarize perinatal and maternal outcomes and resource use for all trials reviewed.

Study Design

All of the included trials were described as "randomized." Four were in fact only pseudorandomized (i.e, treatment was allocated based alternate medical record numbers or birth dates, rather than by randomly generated numbers), which introduces the possibility of bias (Cardozo, Fysh, and Pearce, 1986; Hedén, Ingemarsson, Ahlström, et al., 1991; Katz, Yemini, Lancet, et al., 1983; Ohel, Rahav, Rothbart, et al., 1996). Two studies did not describe the method of randomization used (Egarter, Kofler, Fitz, et al., 1989; Herabutya, Prasertsawat, Tongyai, et al., 1992).

As discussed above and pointed out by Crowley (2000), the practical and ethical difficulties of blinding clinicians to either the target intervention or the results of antepartum testing results in an inherent bias against expectant management. Abnormal antenatal monitoring could influence a clinician's thresholds for performing a cesarean section, either by making the diagnosis of "fetal distress" more likely or by a decreased willingness to augment labor aggressively.

In any trial of planned induction versus expectant management with antepartum testing, a certain proportion of women randomized to planned induction will go into spontaneous labor, while a proportion of women randomized to expectant management will have abnormal antepartum testing results; or, as observed in the Canadian Multicenter Post-term Pregnancy Trial (Hannah, Hannah, Hellmann, et al., 1992), patients or providers may request induction. These subjects are quite correctly analyzed in the groups to which they are randomized, rather than in accordance with the "treatment" received, since the trial is not comparing spontaneous delivery to induction, but instead, management strategies undertaken with the knowledge that some women will deliver spontaneously prior to scheduled induction, and some women will require (or request) induction during expectant management.

Outcome Measurement

All studies reported results for "hard" outcomes such as perinatal mortality and cesarean section rates. Reporting of other outcomes of interest was more variable. Many outcomes are subject to inherent difficulties with reproducibility and bias (e.g., the diagnosis of "fetal distress"), variability in operator preferences and skills (e.g., operative vaginal delivery rates), or are of uncertain long-term clinical significance (e.g., meconium-stained amniotic fluid in the absence of meconium aspiration, or Apgar scores). Other measures, such as patient preferences for different management strategies, longer-term neonatal outcomes, and vaginal and perineal trauma, would be of significant interest to patients, clinicians, and policymakers. We identified one cohort study published in 1991 which showed that patients' preferences for induction versus expectant management changed with advancing gestation: 45 percent of women preferred conservative management at 37 weeks, compared with 31 percent at 41 weeks (Roberts and Young, 1991). Measurement of these preferences in light of data published subsequent to this study, and using methods developed and refined in the past decade, is needed. Detailed measurement of both medical and nonmedical costs is also lacking in the studies reviewed.

Comparability and Generalizability

The gestational age at which interventions were begun, as well as the methods used for induction and monitoring, varied between studies. Because variability in these methods may result in quite different outcomes, caution should be used when comparing outcomes that could possibly be affected by different methods of labor induction (such as cesarean section rates or time spent in labor) or different protocols for fetal monitoring (such as perinatal mortality) between studies. In addition, clinical management decisions may vary between practitioners. Especially in smaller trials, unequal distribution of different practitioners with different preferences and thresholds for management of labor may have resulted in some differences in outcomes.

Readers also must consider the degree to which these studies are generalizable to particular settings. If these methods or protocols are substantially different from those used in a particular setting, then the results may not be applicable. For example, the Canadian Multicenter Post-term Pregnancy Trial did not use prostaglandins for induction of women with abnormal antepartum testing (Crowley, 2000; Hannah, Hannah, Hellmann, et al., 1992). Use of prostaglandins could have changed the results by yielding lower cesarean rates in the induction arm through more successful inductions, as pointed out by Crowley (2000). On the other hand, the use of these agents in women with potentially compromised fetuses could have resulted in even higher cesarean section rates because of fetal compromise. A reanalysis of the Canadian trial using published success rates for prostaglandins found that more liberal use of these agents would still lead to a significantly higher cesarean section rate in the expectant management group because the cesarean section rate in the group induced because of abnormal testing would be substantially higher (Hannah, Huh, Hewson, et al., 1996).

Statistical Issues

Only the Canadian trial (Hannah, Hannah, Hellmann, et al., 1992) was sufficiently powered to detect differences in rare perinatal outcomes. Many of the remaining studies were also under-powered to detect differences in dichotomous outcomes.

Inappropriate summary measures and statistical tests were frequently used (e.g., mean parity or Bishop score, with comparison by t-test, when nonparametric statistics would be more appropriate). Variables that are frequently not normally distributed, such as length of stay and costs, also were not uniformly reported using medians, and the effect of a few outliers on comparisons was not evaluated.

Castor Oil

We identified one randomized trial of castor oil used at term to promote spontaneous labor. Garry, Figueroa, Guillaume, et al. (2000) randomized women to 60 mg castor oil given orally in apple or orange juice (n = 52) or no treatment (n = 48). Mean gestational age was 284.4 ± 4.2 days in the castor oil group and 284.7 ± 3.6 days in the no treatment group. In the castor oil group, 57.7 percent of the subjects were in labor within 24 hours compared with 4.2 percent in the no treatment group (p < 0.001). Cesarean section rates were 19.2 percent in the castor oil group and 8.3 percent in the no treatment group (p = 0.20), but the study was underpowered to detect this difference or differences in rare outcomes such as uterine rupture. Of note, all women in the castor oil group experienced nausea. Other outcomes, such as proportion of women induced for other reasons or neonatal outcomes, were not reported.

The RCOG guideline (Royal College of Obstetricians and Gynaecologists, 2001) did not address castor oil. The most recent Cochrane review on the topic (Kelly, Kavanagh, and Thomas, 2001) identified the article cited above (Garry, Figueroa, Guillaume, et al., 2000) and reached conclusions similar to our own.

Breast Stimulation

We identified two studies that evaluated the use of breast stimulation in promoting the onset of labor near term and one that evaluated breast stimulation as a method of induction. Elliot and Flaherty (1984) randomized 100 women to either breast stimulation (manual stimulation of the nipple and areola for 15 minutes, alternating breasts, for a total of 1 hour at a time, three times daily) beginning at 39 weeks or a control pelvic examination; women in the control group were asked to abstain from sexual intercourse and avoid breast stimulation. Both groups were reevaluated at 42 weeks. Women with Bishop scores of 8 or greater were induced; others were followed with contraction stress tests. Five women in the breast stimulation group reached 42 weeks, compared with 17 in the control group; significance testing was not performed. Women in the breast stimulation group were significantly less likely to be induced after 42 weeks. The study was underpowered to detect differences in important outcomes, especially for the subgroup of women beyond 42 weeks.

Kadar, Tapp, and Wong (1990) randomized women at 39 weeks to either daily unilateral manual nipple stimulation "for as long as was practically feasible" (n = 60) or to no nipple stimulation (n = 76). There were no significant differences in any of the outcomes reported, including the proportion going into spontaneous labor, postterm deliveries, or median duration of pregnancy. Survival analysis showed that duration of pregnancy was related only to gestational age at enrollment and Bishop score. The authors also noted that adherence to the prescribed regimen was poor: 70 percent of the women assigned to the nipple stimulation group either failed to perform nipple stimulation at all or did so for less than 2 hours total during the entire study.

Chayen, et al., compared nipple stimulation using an electric breast pump to oxytocin as a method of induction (Chayen, Tejani, and Verma, 1986). In this study, only 29 percent of the inductions were for prolonged pregnancy. Thirty subjects were induced initially with a breast pump, while 32 received oxytocin. Time to achieve regular contractions and adequate labor as documented by intrauterine catheter were significantly less in the breast pump group. Cesarean section rates were also lower (26.7 percent vs. 43.7 percent in the oxytocin group), although this difference was not significant. Patients in the oxytocin group were more likely to have a higher Bishop score at baseline. Results were not reported separately by parity or for the subgroup of women induced for prolonged pregnancy.

In summary, because of lack of significance testing, poor compliance, or lack of power, the available randomized trials do not allow conclusions to be drawn about the effectiveness of breast stimulation in promoting labor or as a method of induction. The RCOG guideline (Royal College of Obstetricians and Gynaecologists, 2001) did not address this topic.

Relaxin

We identified three randomized trials of relaxin. Evans, Dougan, Moawad, et al. (1983) randomized women at 41 weeks gestation scheduled to undergo oxytocin induction of labor to intracervical or vaginal insertion of 4 mg relaxin (n = 10), 2 mg relaxin (n = 13), or placebo (n = 14); if the patient reached 42 weeks gestation, then labor was induced. No significant differences in any parameters, including days to admission, spontaneous labor, or time to delivery, were noted. There were trends towards a shorter time to delivery in the relaxin groups, but the study was underpowered to detect a difference for this outcome.

Bell, Permezel, MacLennan, et al. (1993) randomized women scheduled for induction for prolonged pregnancy to intravaginal 1.5 mg recombinant human relaxin (n = 18) or placebo (n = 22). No significant differences in any outcomes were reported. The authors noted that a low dose was deliberately chosen to help establish a safety profile for relaxin.

Brennand, et al., randomized women between 37 and 42 weeks, "most" of whom were being induced for pregnancy-induced hypertension or prolonged pregnancy, to placebo or 1 mg, 2 mg, or 4 mg of recombinant relaxin (Brennand, Calder, Leitch, et al., 1997). There were no significant differences in any outcome except for slightly elevated baseline fetal heart rates after relaxin.

In summary, there are insufficient data available on relaxin to draw any conclusions about its safety or efficacy in induction of labor in women with prolonged pregnancy.

Sweeping of the Membranes

Table 27: Sweeping of the membranes to promote labor (more...)

Table 27: Sweeping of the membranes to promote labor and reduce need for induction

Study	Membrane stripping	Comparison	N	Statistically Significant Differences
Idrisa, Obisesan, and Adeleye, 1993	41 weeks, once	Not described	200	Spontaneous labor within one week: 92% sweeping vs. 33% control
Magann, McNamara, Whitworth, et al., 1998	39 weeks, stripping every 3 days	Vaginal examination	65	Inductions at 42 weeks: 0 stripping vs. 56% controls
Salamalekis, Vitoratos, Kassanos, et al., 2000	40-41 weeks, membrane stripping once	6-hour oxytocin infusion or cervical examination	104	Spontaneous labor: 68% stripping, 51% oxytocin vs. 34% control Induction of labor: 2.9% stripping, 5.7% oxytocin vs. 20% control
Gupta, Vasishta, Sawhney, et al., 1998	38 weeks, not specified if repeated	Cervical examination	100	Mean gestational age at delivery: 38.7 weeks stripping vs. 39.8 control Mean days from randomization to delivery: 4.6 stripping vs. 11.9 control Pregnancies beyond 40 weeks: 4% stripping vs. 34% control Induction of labor: 2% stripping vs. 32% control
Wiriyasirivaj, Vutyavanich, and Ruangsri, 1996	38 weeks, repeated weekly	Cervical examination	120	Proportion delivering with 7 days: 41% stripping vs. 20.3% control
Cammu and Haitsma, 1998	39 weeks, repeated weekly	"Routine pelvic examination"	278	Proportion reaching 41 weeks: 19% stripping vs. 33% control Induction of labor: 11% stripping vs. 26% control
Berghella, Rogers, and Lescale, 1996	38 weeks, repeated weekly	"Gentle cervical examination"	139	Delivery after 41 weeks: 5% stripping vs. 22% control Mean days to delivery: 8.2 days stripping vs. 12.2 days control
McColgin, Hampton, McCaul et al., 1990	38 weeks, repeated weekly	"Atraumatic assessment of cervix"	180	Delivery beyond 42 weeks: 3.3% stripping vs. 15.6% control Mean days to delivery: 8.6 days stripping vs. 15.1 days control Delivery within 1 week: 54.5% stripping vs. 15.6% control
Allott and Palmer, 1993	40+ weeks, one time	Cervical examination for Bishop score	195	Induction of labor: 8.1% stripping vs. 18.8% control Time to delivery: 2.24 days stripping vs. 5.18 days control (No differences in primigravidas with high Bishop score)
El-Torkey and Grant. 1992	41-42 weeks, twice weekly	No vaginal examination	65	Spontaneous labor prior to scheduled induction at 42 weeks: 76% stripping vs. 38% control
Crane, Bennett, Young, et al., 1997	38-40 weeks, once	Cervical examination	150	Epidural anesthesia: 66% stripping vs. 43% control Premature rupture of membranes: 6.6% stripping vs. 22% control
McColgin, Patrissi, and Morrison, 1990	38-42 weeks, repeated weekly	Cervical examination for Bishop score	99	Days to delivery: 6.7 stripping vs. 13.3 control Proportion delivering after 1 week: 59% stripping vs. 21% control

All studies except one consistently showed higher rates of labor within a predefined time period, usually 1 week, in women randomized to active membrane sweeping. The proportion of women induced was also consistently lower in groups randomized to membrane sweeping. No differences in adverse outcomes, including infection or bleeding, were noted in any study. Level of patient discomfort during the procedure was not assessed in any study.

The one study that did not show a difference in outcomes (Crane, Bennett, Young, et al., 1997) was different from the other trials in several ways. Membrane stripping was performed only once. Patients in the stripping group were more likely to be nulliparous and to have lower Bishop scores. Stratified analyses and logistic regression did not show significant effects, but it is possible that the smaller sample size in these subgroups limited power. In addition, a survival analysis showed a decrease in the median time from enrollment to delivery (6.5 days for stripping, compared with 8 days for controls), but this difference was not significant.

In the one study in which membrane sweeping was used as an adjunct to induction of labor, Boulvain, et al., randomized women to sweeping of the membranes (n = 99) or vaginal examination only (n = 99) prior to induction of labor for "nonurgent" indications (Boulvain, Fraser, Marcoux, et al., 1998). Eighty-five percent of the patient population was induced for prolonged pregnancy. Mean time from randomization to onset of labor was significantly shorter in the sweeping group (76 hours vs. 98 hours; p = 0.01), but no significant differences were seen in other outcomes except patient discomfort (odds ratio [stripping vs. control], 2.52; 95 percent confidence interval [CI], 1.60 to 3.99), bleeding, and painful contractions without labor.

In summary, in all but one study, sweeping the membranes consistently promoted labor at term and reduced the incidence of induction for prolonged pregnancy. As with the majority of the interventions reviewed in this report, there are no data on patient preferences for this intervention. One study found that women who undergo membrane stripping are more likely to experience discomfort, bleeding, and painful contractions without labor compared with controls. Another issue is that the majority of studies excluded women whose cervices would not allow introduction of the examiner's finger; thus, the conclusions described are applicable only to those pregnant women at term whose cervices are dilated enough to allow introduction of an examiner's finger.

Similar findings have been reported in a Cochrane review (Boulvain and Irion, 2001) and incorporated into the RCOG guidelines (Royal College of Obstetricians and Gynaecologists, 2001).

Mechanical Devices

We identified two randomized trials of the use of mechanical devices such as Foley catheters, which are inserted into the cervix and then inflated. Atad, et al. (Atad, Hallak, Auslender, et al., 1996) compared 3 mg PGE₂ gel (n = 30), oxytocin (n = 30), and a double-balloon catheter invented by one of the investigators (n = 35). Patients in the first two groups crossed over to the catheter arm if the Bishop score was < 4 at 12 hours, while patients in the catheter group received PGE₂ if the Bishop score was < 4 at 12 hours. More patients in the catheter group had cervical dilation > 3 cm after 12 hours (86 percent vs. 23 percent in the oxytocin group and 50 percent in the PGE₂ group; p < 0.01). Both PGE₂ and the balloon device had higher rates of vaginal delivery (PGE₂, 70 percent; catheter, 77 percent; oxytocin, 27 percent) and lower rates of cesarean section among patients with cervical dilation after the initial intervention (PGE₂, 13 percent; catheter, 18 percent; oxytocin, 43 percent). Only 18 percent of the inductions in this study were for prolonged pregnancy.

Sciscione, et al., randomized 53 women to misoprostol and 58 to mechanical dilation with a 16 F Foley catheter with a 30 cc balloon (Sciscione, Nguyen, Manley, et al., 2001). There were no significant differences in change in Bishop score, vaginal delivery rates, or time to delivery in the two groups. Uterine tachysystole and passage of meconium were significantly more frequent in the misoprostol group. There was a trend towards higher cesarean section rates for nonreassuring fetal heart rate tracing in the misoprostol group (24 percent vs. 12 percent; p = 0.09), in a study where the sample size was determined based on change in Bishop score. Only 16 of 111 women in this study were induced for an indication of prolonged pregnancy.

In these two trials, mechanical devices appear to be comparable to prostaglandins in terms of delivery success, with lower rates of fetal heart rate tracing changes associated with frequent uterine contractions. As with membrane sweeping, applicability is limited to women whose cervix is dilated enough to allow introduction of a catheter. As with the majority of the other interventions reviewed, these studies also included relatively few women in the population of interest (prolonged pregnancy with no other risk factors) and were underpowered to detect differences in many important outcomes.

Mechanical devices alone are not addressed specifically in published Cochrane reviews or in the RCOG guideline (Royal College of Obstetricians and Gynaecologists, 2001).

Oyxtocin Dosing

We identified one randomized trial comparing two dosing regimens of oxytocin. Satin, Hankins, and Yeomans (1991) randomized women being induced for prolonged pregnancy to a "slow-dose" regimen (an initial dose of 2 mU/min, with increments of 1 mU/min at 30-minute intervals) or a "fast-dose" regimen (an initial dose of 2 mU minute with increases of 2 mU/min at 15-minute intervals). Induction failure was more likely in the slow-dose group (31 percent vs. 8 percent; p < 0.05). Time to delivery was shorter in the fast-dose group in both nulliparous women (9 hours vs. 15 hours; p < 0.05) and multiparous women (8 hours vs. 11 hours; p < 0.05). No significant differences were observed in other outcomes. There was a trend towards more hyperstimulation episodes requiring cessation of oxytocin in the fast-dose group, but the study was underpowered to detect a difference.

There is no formal comparison of oxytocin dosing regimens in published Cochrane reviews. The RCOG guideline development group reviewed dosing regimens in 11 trials of oxytocin with and without amniotomy. Their qualitative conclusions were: (1) lower dose regimens were not associated with an increase in operative delivery rates; (2) regimens with incremental rises in dose more frequently than every 30 minutes were associated with an increase in uterine hypercontractility; (3) lower dose regimens were not associated with an increase in specified delivery intervals; and 4) higher dose regimens were associated with an increase in the incidence of precipitous labor (Royal College of Obstetricians and Gynaecologists, 2001).

Prostaglandins

Misoprostol

Misoprostol tablets versus PGE₂ gel

Table 28: Misoprostol tablets versus PGE2 gel

Table 28: Misoprostol tablets versus PGE2 gel

Study	Misoprostol Dose	PGE2 Dose	N	% Postterm	Statistically Significant Differences
Buser, Mora, and Arias, 1997	50 μg, posterior fornix	0.5 mg, intracervical	155	35%	Nonreassuring FHR tracing: 14% misoprostol vs. 0% PGE2 Time to delivery: 15.8 hours misoprostol vs. 24.2 hours PGE2 C-section for nonreassuring FHR: 25% misoprostol vs. 5% PGE2
Chuck and Huffaker, 1995	50 μg, posterior fornix	0.5 mg, intracervical	99	18%	Vaginal delivery with 24 hours: 100% misoprostol vs. 68% PGE2 Time to vaginal delivery: 11.4 hours misoprostol vs. 18.9 hours PGE2
Fletcher, Mitchell, Frederick, et al., 1994	100 μg, posterior fornix	3 mg, posterior fornix	63	33%	None
Gottschall, Borgida, Mihalek, et al., 1997	100 μg, posterior fornix	5 mg, posterior fornix	75	40%	Time to vaginal delivery: 31.3 hours misoprostol vs. 58.5 hours PGE2
Herabutya, Prasertsawat, and Pokpirom, 1992	100 μg, posterior fornix	1.5 mg, intracervical	110	34%	None
Howarth, Funk, Steytler, et al., 1996	100 μg, posterior fornix	1 mg, posterior fornix	72	42% misoprostol, 25% PGE2	C-sections: 17% misoprostol vs. 42% PGE2 Delivery within 12 hours: 83% misoprostol vs. 36% PGE2 Tachysystole: 39% misoprostol vs. 8% PGE2
Kadanali, Küçüközkan, Zor, et al., 1996	100 μg, posterior fornix	Dose not specified	224	41%	Time to delivery: 9.2 hours misoprostol vs. 15.2 PGE2 Cesarean section for failed induction: 6.3% misoprostol vs. 13.4% PGE2
Mundle and Young, 1996	50 μg, upper vagina	0.5 mg, intracervical or 1-2 mg, intravaginal	222	78%	Time from induction to delivery: 753 minutes misoprostol vs. 941 minutes PGE2
Varaklis, Gumina, and Stubblefield, 1995	25 μg, intravaginal	0.5 mg, intracervical	69	Not stated	Time to vaginal delivery: 15.7 hours misoprostol vs. 20.7 hours PGE2
Wing, Jones, Rahall, et al., 1995	50 μg, posterior fornix	0.5 mg, intracervical	135	10%	Time to vaginal delivery: 903 minutes misoprostol vs. 1411 minutes PGE2 Vaginal delivery within 24 hours: 70.6% misoprostol vs. 47.8% PGE2 Tachysystole: 36.7% misoprostol vs. 11.9% PGE2 Need for neonatal resuscitation: 22.1% misoprostol vs. 7.5% PGE2 Meconium aspiration: 4.4% misoprostol vs. 1.5% PGE2
Wing, Rahall, Jones, et al., 1995	25 μg, posterior fornix	0.5 mg, intracervical	275	16%	Time to vaginal delivery: 1323 minutes misoprostol vs. 1532 minutes PGE2 Vaginal delivery with 24 hours: 65.5% misoprostol vs. 41.4% PGE2

Table 28 summarizes results from the 10 studies that compared intravaginal misoprostol tablets with intracervical or intravaginal PGE₂ gel (Buser, Mora, and Arias, 1997; Chuck and Huffaker, 1995; Fletcher, Mitchell, Frederick, et al., 1994; Gottschall, Borgida, Mihalek, et al., 1997; Herabutya, Prasertsawat, and Pokpirom, 1997; Howarth, Funk, Steytler, et al., 1996; Kadanali, Küçüközkan, Zor, et al., 1996; Mundle and Young, 1996; Varaklis, Gumina, and Stubblefield, 1995; Wing, Jones, Rahall, et al., 1995).

The studies examined a range of doses and frequency of dosing with similar results. The time from induction to delivery was consistently shorter in patients treated with misoprostol, both for all patients and for those with vaginal delivery. With one exception, misoprostol was shown to cause higher frequency of uterine hyperstimulation, hypertonus, or tachysystole, although studies were often underpowered to detect significant differences in these outcomes. All studies indicated that misoprostol was an effective agent for cervical ripening and induction, often more effective than PGE₂ gel, and showed no significant difference in the rates of cesarean section. One study (Buser, Mora, and Arias, 1997) showed an increase in cesarean section rates for patients treated with misoprostol; this was attributable to significantly higher rates of nonreassuring fetal heart rate patterns. Of note, the majority of subjects in these studies were not women being induced for prolonged pregnancy.

Mifepristone

We identified five studies that compared the efficacy of the progesterone receptor antagonist mifepristone (RU-486) to placebo. Unlike many of the studies discussed above, three of the five focused on patients primarily induced for prolonged pregnancy. All five studies indicated that mifepristone was effective in ripening the cervix. Wing, et al., using 200 mg mifepristone, found significantly more deliveries and vaginal deliveries within 48 hours and a shorter time to delivery with mifepristone compared with placebo; subgroup analysis showed that these effects were primarily due to the effect in nulliparas (Wing, Fassett, and Mishell, 2000). There were trends towards more abnormal fetal heart rate tracings in labor and more infants with Apgar scores less than 7 at 1 and 5 minutes in the mifepristone group, but these trends did not reach statistical significance.

Three studies evaluated patients who were treated with 400 mg mifepristone versus placebo. In Stenlund, Ekman, Aedo, et al. (1999), the time to onset of labor was shorter and the proportion of patients in labor within 48 hours was significantly greater (81.8 percent vs. 27.3 percent) in the mifepristone group. Median Apgar scores at 1 minute were lower in the mifepristone group, but there were no differences in Apgar scores at 5 or 10 minutes. With only 36 subjects, this study was underpowered to detect differences in many outcomes.

In Giacalone, et al., time to onset of labor and time to vaginal delivery were significantly shorter in the mifepristone group (Giacalone, Targosz, Laffargue, et al., 1998). There were trends towards lower Apgar scores at 1 minute and lower cord pH values, but these were nonsignificant; again, the study was severely underpowered to detect differences in many important clinical outcomes, including cesarean section rate.

In Frydman, et al., the proportion of women going into spontaneous labor, the proportion with Bishop scores less than 4 at presentation for induction, and the mean randomization-to-delivery time were all significantly less in the mifepristone group (Frydman, Lelaidier, Baton-Saint-Mleux, et al., 1992). There were no significant differences in other outcomes and no other trends. Again, the study was underpowered to detect differences in safety-related outcomes. Forty-eight percent of the patients were induced for "postdate" pregnancy.

Elliott, et al., performed a dose-response study comparing placebo with 50 mg and 200 mg of mifepristone in nulliparous women, the "majority" of whom were being induced for prolonged pregnancy (Elliott, Brennand, and Calder, 1998). When a combined outcome measure of either spontaneous labor within 4 days or Bishop score of > 6 at induction was used as the measure of efficacy, there were significant improvements with mifepristone in a dose-related manner. However, mifepristone was also associated in a dose-related manner with significantly more cases of fetal distress in labor and neonatal jaundice. In addition, cesarean rates were significantly lower with 50 mg of mifepristone than with placebo but higher with 200 mg than with placebo (p = 0.07), a difference that appears to be attributable to a higher incidence of cesarean delivery for fetal distress in the 200-mg group.

In summary, mifepristone appears to be superior to placebo in terms of achieving labor or cervical ripening within a specified time, but there are consistent trends towards fetal compromise during labor in women who receive mifepristone. Inadequate power to detect potentially important differences in safety argue against the use of mifepristone for induction of labor in prolonged pregnancy outside of research protocols at the present time.

A Cochrane review on this topic found similar evidence of efficacy (Neilson, 2001). Neonatal outcomes were not reported in enough studies to allow conclusions about safety.

Racial and Ethnic Differences: Literature Review

We did not identify any articles that specifically addressed differences in the epidemiology or outcomes of prolonged pregnancy in different ethnic groups.

Racial and Ethnic Differences: Primary Data

Birth certificate data

Table 29: Births by race of mother (1998 birth certificate (more...)

Table 29: Births by race of mother (1998 birth certificate data)

Race	Total births	% ≥ 40 weeks	% ≥ 41 weeks	% ≥ 42 weeks	% ≥ 42 weeks and < 2,500 g	% ≥ 42 weeks and > 4,000 g
Total	3,941,553	39.6%	18.7%	7.4%	2.2%	14.6%
White, non-Hispanic	2,361,462	41.8%	19.4%	7.5%	1.8%	16.8%
Black, non-Hispanic	593,127	35.3%	16.6%	7.2%	4.0%	8.0%
Hispanic	734,661	40.2%	18.7%	7.7%	2.0%	12.8%

Table 29 summarizes total births, with percentages of infants born after 40 weeks, 41 weeks, and 42 weeks, from 1998 birth certificate data reported to the National Center for Health Statistics (NCHS), by race of mother (Asian or Native American data are not available in the published report). The proportions reported were calculated from the absolute numbers provided in the NCHS report. Table 29 also illustrates the proportion of live births after 42 weeks that were low birthweight (less than 2,500 grams) or macrosomic (greater than 4,000 grams).

Taking into account the limitations of birth certificate data, there are some interesting findings:

• Live births between 40 and 42 weeks were less common for non-Hispanic black women than for non-Hispanic white women, which may be partly due to an increased risk of preterm birth among non-Hispanic blacks (17.5 percent vs. 10.2 percent in non-Hispanic whites). However, the proportion of births after 42 weeks is strikingly similar in all groups.

• The weight distribution among infants born after 42 weeks is also strikingly different between groups, with non-Hispanic black women having a two-fold increase in low birthweight infants and a substantially lower incidence of macrosomic infants.

Hospital discharge data

Table 30: Percent distribution of selected secondary (more...)

Table 30: Percent distribution of selected secondary discharge diagnoses, by race, in patients with discharge diagnosis of "prolonged pregnancy"

Race	None	IUFD	IUGR	Fetal distress	Oligo-hydramnios	Macrosomia	Shoulder dystocia	Perineal trauma	Labor abnormalities	Failed induction
White	36.38%	0.02%	0.38%	8.55%	2.30%	3.80%	1.86%	28.26%	14.57%	3.87%
Black	35.92%	0.00%	0.71%	14.34%	4.20%	2.73%	2.02%	21.29%	14.91%	3.87%
Hispanic	37.63%	0.02%	0.36%	11.39%	3.83%	3.77%	1.25%	23.28%	13.67%	4.79%
Asian/Pacific Islander	30.34%	0.00%	0.64%	7.50%	4.27%	8.91%	0.81%	24.64%	19.65%	3.23%
Native American	40.62%	0.00%	0.00%	10.92%	5.60%	2.80%	0.00%	25.49%	11.48%	3.08%
Other	33.53%	0.00%	0.48%	10.15%	3.10%	2.61%	1.90%	29.52%	14.24%	4.47%

IUFD = Intrauterine fetal demise

IUGR = Intrauterine growth restriction

Table 30 shows the percentage distribution of selected discharge diagnoses in the subset of women with a primary discharge diagnosis of prolonged pregnancy, by coded ethnic group. Total raw discharges in the NIS with this diagnosis were 57,814, or 7.2 percent of the total pregnancy-related discharges. Again, black women were more likely than women in other ethnic groups to have a diagnosis of restricted fetal growth and were less likely to have a diagnosis of macrosomia than white or Hispanic women. Black women also were more likely to have diagnoses of fetal distress and oligohydramnios. Interestingly, they also were somewhat more likely to have a diagnosis of shoulder dystocia than white or Hispanic women. Asian/Pacific Islander women were more likely to have diagnoses of macrosomia but less likely to have perineal trauma of any kind. Potential explanations for this observation include a higher cesarean section rate in Asian/Pacific Islander women, differences in the pelvic floor, or dynamics of labor which make perineal trauma less likely.

Both the NIS data and birth certificate data suggest that black women are more likely to have low birthweight infants after 42 weeks than white or Hispanic women. Diagnoses such as oligohydramnios and fetal growth restriction are also more common in black women. All three of these diagnoses are consistent with declining uteroplacental function. There were a limited number of fetal deaths in the NIS data set, with racial data missing from over half.

Socioeconomic Groups: Literature Review

We did not identify any articles that specifically addressed differences in the epidemiology or outcomes of prolonged pregnancy in different socioeconomic groups.

Socioeconomic Groups: Primary Data

Table 31: Percent distribution of selected secondary (more...)

Table 31: Percent distribution of selected secondary discharge diagnoses, by payer, in patients with discharge diagnosis of "prolonged pregnancy"

Payer	None	IUFD	IUGR	Fetal distress	Oligo-hydramnios	Macrosomia	Shoulder dystocia	Perineal trauma	Labor abnormalities	Failed induction
Medicare	37.14%	0.00%	0.00%	13.69%	0.83%	5.60%	0.00%	22.61%	17.22%	2.90%
Medicaid	37.53%	0.06%	0.58%	10.50%	3.46%	2.79%	1.66%	25.75%	13.77%	3.90%
Private/HMO	35.05%	0.01%	0.34%	9.47%	2.43%	4.24%	1.80%	27.82%	15.13%	3.72%
Self-pay/No insurance	40.22%	0.00%	0.61%	10.16%	4.18%	2.47%	1.90%	26.71%	11.02%	2.73%
"No charge"	40.72%	0.00%	0.00%	0.52%	1.55%	6.19%	0.00%	33.51%	7.73%	9.79%
Other	37.11%	0.00%	0.73%	10.64%	1.86%	3.75%	2.09%	27.66%	12.73%	3.43%

IUFD = Intrauterine fetal demise

IUGR = Intrauterine growth restriction

Age Differences: Literature Review

We did not identify any articles that specifically addressed differences in the epidemiology or outcomes of prolonged pregnancy in either adolescent women or women in their later reproductive years.

Data Quality Issues

The accuracy of the dating recorded on birth certificates is unconfirmable, at best. Therefore, it is unclear whether the observed trends in racial differences in the distribution of birthweight after 42 weeks, and the observed lack of difference in the proportion of all pregnancies that reach 42 weeks, are real or simply random error introduced by variable quality of dating.

Similarly, criteria for a diagnosis of prolonged pregnancy, as well as for many of the other diagnosis codes, may vary between hospitals. Data for racial and payer codes were missing for many of the coded complication diagnoses. If codes are not recorded systematically in some hospitals, this may result in misleading patterns.

Statistical Analysis

Because of concerns with data quality, we did not perform formal tests of significance or multivariate analyses. Given the consistent patterns for some observations seen in the two data sets, more detailed analysis of more complete data sets is warranted.

Literature Search

We used standard methods for identifying, reviewing, and abstracting published studies focused on the management of prolonged pregnancy. We used predefined study characteristics to identify those studies most likely to provide unbiased estimates of efficacy and test performance. We did not search the literature prior to 1980, primarily because we assumed that the lack of general availability of ultrasound for both dating and management of prolonged gestation would limit the applicability of these results to current practice. We also limited our search to articles published in English, primarily for reasons of convenience and resource constraints. It is possible that including older studies, or studies published in other languages, would have identified additional evidence that would have substantially changed our conclusions. This may be especially true for alternative or complementary therapies.

Another limitation of our exclusion criteria is that rare but severe complications of treatments may have been overlooked because they were published in case reports or small case series. Although these study designs are useful for identifying potential problems, it is difficult to quantify these risks when only numerator values are available.

Grading of Articles

We did not use one of the currently available quality scoring systems to grade the articles we reviewed. However, we believe that the rationale for each criterion we used is reasonable, and that the operational definitions are clear and reproducible. In addition, we used these grading criteria primarily to provide additional detail to other researchers. We did not use them to establish a threshold for including or excluding articles or to weight the results of a quantitative evidence synthesis such as a meta-analysis.

Other Data Sources

We used one additional data source in preparing this report, the Nationwide Inpatient Sample (NIS) (Nationwide Inpatient Sample [NIS], 1997). The NIS, like most administrative databases, is limited by a lack of clinically relevant detail. In addition, even the data recorded in these discharge abstracts were incomplete, limiting our ability to analyze them in great detail. Variability in definitions between hospitals also may lead to incorrect conclusions. The primary value of these data in the context of this report is to identify potentially important differences in outcomes between ethnic and socioeconomic groups that need to be explored further in data sets with better documentation and more complete data.

Perinatal Mortality

The most precise data available come from the United Kingdom. Estimates in U.S. populations, preferably with the ability to control for the presence of other risk factors for mortality and the use of antepartum testing, are needed. Potential studies include:

• Detailed analysis of U.S. birth certificate data.

• Detailed analysis of U.S. hospital discharge data, although this will necessarily miss deliveries performed outside the hospital, such as those performed at freestanding birth centers and home births.

• Detailed analysis of administrative or computerized clinical data from large provider organizations, such as health maintenance organizations.

Because of the inherent limitations of these data sources, validation with detailed clinical records ultimately will be needed to systematically determine and describe causes of death. These data also would allow determination of the impact of various methods of dating pregnancy on perinatal mortality.

Perinatal Morbidity

Similar methods need to be applied to estimations of the risks of perinatal morbidity:

• Careful attention should be given to case definitions; again, validation of the accuracy of administrative data is needed.

• We did not identify any recent publications providing followup data on infants born after prolonged gestation. Ultimately, long-term outcomes are most important, and better data on the long-term consequences of various management strategies are needed.

Maternal Morbidity

• Again, better estimation of the risks, given current obstetric practice, is needed.

• Recently, attention has been drawn to the risks of long-term maternal consequences of labor and delivery, especially pelvic floor dysfunction. It is unclear if any of the management strategies used for prolonged pregnancy have any impact on the risks of subsequent development of pelvic floor dysfunction.

POST-TERM PREGNANCYARTICLE ABSTRACTING FORM

Reviewer:_________________First Author:___________________________Year:___________Procite #:___________

ARTICLE FOCUS (circle one): Testing / Management / Both

STUDY DESIGN (check one):

______Cohort

______Case series, no controls, n = ______

______Case series, historical controls, n = ______

______Case series, concomitant controls, n = ______

______Not specified or unable to classify

REASSESSMENT:

KEY QUESTIONS ADDRESSED (check all that apply):

_____1. What are the test characteristics (reliability, sensitivity, specificity, predictive values) and costs of measures used in the management of postdates pregnancy: (a) to assess risks to the fetus of postdates pregnancy, and (b) to assess the likelihood of a successful induction?

_____2. What are the benefits, risks, and costs of currently available interventions for induction of labor?

_____3. What is the direct evidence comparing the benefits, risks, and costs of planned induction versus expectant management at various gestational ages?

_____4. Are the epidemiology and outcomes of postdates pregnancy different for women in different ethnic groups, different socioeconomic groups, or in adolescent women?

STUDY LOGISTICS:

Inclusive dates of data collection (give month and year): from________________________to_____________________

Multicenter study? (circle one): Yes / No If "Yes," no. of sites:_________

Geographic location (in US, give city and state; outside of US, give city and country. If multicenter trial or network, give name, e.g., NICHD MFM Network, RADIUS):___________________________________________________

GESTATIONAL AGE DETERMINED BY (check all that apply):

______LMP

______1^st trimester U/S

______2^nd trimester U/S

______Other - specify: __________________________________________________________________________

SUBJECT CHARACTERISTICS:

1. Identify interventions A, B, and C, and indicate which (if any) served as control

2. Use "NR" to indicate "Not reported"

INTERVENTIONS

Describe the testing and management interventions used in each study group. Include all information necessary to reproduce the treatment/monitoring/testing algorithms used. For example:

Interventions to be considered include:

1. Tests of fetal well-being: No tests, nonstress test, biophysical profile, contraction stress test, amniotic fluid volume, uterine vessel Doppler flow, other, combinations of the preceding

2. Tests of fetal size: Physical exam, ultrasound, other

3. Tests of readiness for delivery: Bishop score, fetal fibronectin, other, combinations of the preceding

4. Interventions: Monitoring/conservative care, stripping of membranes, oxytocin, prostaglandin gel, misoprostil, mechanical interventions

PATIENT NUMBERS, DROPOUTS AND LOSS TO FOLLOW-UP:

MANAGEMENT OUTCOMES:

TEST PERFORMANCE OUTCOMES (Testing Articles Only):

COST/CHARGES/RESOURCE UTILIZATION OUTCOMES:

QUALITY SCORE:

(Check "Yes" or "No" for each item)