From obstetric ultrasound to ultrasonographic obstetrics
Obstetrics has long been the most primitive form of clinical medicine, based largely on the observation of dramatic situations and management with ancient maneuvers. Slowly rising above this, obstetricians started to care about the fetal status; first with a stethoscope and then with its electrical counterpart, the cardiotocogram. It was not that long ago that we became able to access the womb with an ultrasound view. Since then, obstetrics has been channeled into the diagnosis of adverse fetal conditions, and diverted from its origins in the labor ward. When an ultrasound machine first arrived in the perinatal unit is hard to tell but, since that day, obstetrics has been gradually taken over by ultrasonography1.
Evidence or relevance?
Randomized controlled trials (RCTs) were designed to evaluate the effectiveness of medical intervention and are therefore only indirectly applicable to investigation of obstetric ultrasonography. Most ultrasound-based RCTs are over 15 years old and report on practices that are older still, with some of the more recent ones inappropriate for dealing with prenatal ultrasonography since mortality and severe morbidity are unlikely in preselected low-risk pregnancies2. On an individual basis, and owing to the simplicity of this largely innocuous technique, it is unfair to pronounce ultrasonography useless in unselected populations. As rare as adverse obstetric events are in unselected pregnancies, they are at least 100 times more frequent than the most common fetal abnormality in the same population. Clinically accepted indications for obstetric ultrasonography outside the diagnosis of fetal anomalies have reached 27 that can be spelled out in twice as many different clinical situations3. The likelihood of any pregnancy slipping through this net is probably closer to 5% than 95%, therefore at odds with the ‘magic numbers’ used in antenatal screening.
A needle in a haystack
The sensitivity of ultrasonography is usually good when performed by a competent operator subjected to quality control. However, its predictive value depends upon the prevalence of the adverse event under investigation. As an example, observation of a short cervix together with painful contractions now defines the diagnosis of true preterm labor4. Cervical length of less than 20 mm at mid-gestation in an asymptomatic woman with no other risk factors predicts preterm delivery in up to only 26% of cases. However, the negative predictive value of ultrasonography is very high (96.5%) and its ability to give reassurance is often played down as compared to the potentially unnecessary interventions generated by a false-positive result. Ultrasound examination is mostly used to assess the indication for a medical intervention, one that unfortunately often turns out to be either inappropriate or unable to prevent the adverse perinatal outcome.
Go wide early
Routine ultrasonography before 24 weeks' gestation is recommended for gestational age assessment and the detection of multiple pregnancies and clinically unsuspected fetal malformations5. Redefining an early scan as one carried out at 11–14 weeks' gestation has improved dating and facilitated the classification of multiple pregnancy types6, 7. This early ultrasound dating has led to a reduction in the incidence of post-term pregnancy (odds ratio (OR), 0.68; 95% confidence interval (CI), 0.57–0.82). As a result of accurate dating, routine induction of labor after 41 weeks' gestation has reduced perinatal mortality (OR, 0.20; 95% CI, 0.06–0.70) without any effect on the Cesarean section rate8.
Early pregnancy failure is also best managed by ultrasonography, which can reliably triage between missed miscarriage and anembryonic pregnancy, where intervention is advisable, and incomplete miscarriage that can justify expectant management in some cases9. Furthermore, ultrasonography has revolutionized the management of ectopic pregnancy, which is no longer the archetypal indication for an emergency surgical procedure but now largely involves elective surgery or medical therapy10.
Most studies on prenatal ultrasonography have been based on a single examination performed at 17–20 weeks' gestation—not appropriate to be called an early scan and too early to have any impact on perinatal outcome. In addition to the advantages of an early scan at 11–14 weeks' gestation, there are distinct advantages of performing a second ultrasound examination at 22–24 weeks11, when the false-positive rates for obstetric targets such as uterine artery Doppler velocimetry, fetal growth assessment and cervical length are expected to be lower4, 12.
Ultrasound assessment of the cervix has proven better than digital examination at predicting preterm labor4. Dilatation of the internal os and sacculation or prolapse of the membranes into the cervix, with shortening of the functional cervical length either spontaneously or induced by transfundal pressure, have provided an ultrasonographic definition of cervical incompetence13. Although evidence is still lacking, there does appear to be benefit in performing a cerclage rather than continuing with expectant management in cases with sonographic appearance of cervical incompetence in asymptomatic women at high risk of preterm delivery. Ultrasonography can also be offered to reduce the indications of cerclage for cases with a positive history in which the situation is uncertain and when cervical length remains stable on serial ultrasound assessment. Similarly, transvaginal sonography might help in the selection of asymptomatic but high-risk women within the general obstetric population. However, the benefit associated with cerclage performed for sonographic indications in this group has not been demonstrated. A more appropriate intervention in these cases could be maintenance treatment with progesterone—subject to the results of several ongoing trials—which would then justify systematic ultrasonographic assessment of cervical length and funneling in all pregnancies14.
The later the better
Until recently, routine ultrasound after 24 weeks' gestation in low-risk or unselected populations had not been shown to affect antenatal, obstetric or neonatal intervention or morbidity in screened vs. control groups, although placental grading was associated with a significant reduction in the stillbirth rate in the one trial that assessed it15. The introduction of two biophysical ultrasound examinations, at 30–32 and 36–37 weeks' gestation, assessing placental maturity, amniotic fluid volume and estimated fetal weight in a low-risk antenatal population showed that the proportion of infants assessed as small for dates at birth was smaller in the study group (6.9% vs. 10.4%; P = 0.008), with a corresponding increase in obstetric interventions (31.3% vs. 16.9%; P < 0.001)16.
Cohort studies show that the use of customized birth-weight centiles improves the detection of small babies at risk of morbidity and mortality when compared with the use of population centiles17, 18. In addition to biometric assessment, there is an established benefit of using Doppler ultrasonography in pregnancies at high risk of growth restriction or pre-eclampsia19. Its use is associated with a reduction in perinatal deaths (OR, 0.71; 95% CI, 0.50–1.01) with fewer inductions of labor (OR, 0.83; 95% CI, 0.74–0.93) and fewer admissions to hospital (OR, 0.56; 95% CI, 0.43–0.72), with no reports of any adverse effects20. The rate of avoidable perinatal deaths is also lower (20%) in cases monitored by Doppler ultrasonography than in controls (50%)21.
An example of clinical intervention which could, at a glance, reflect poorly on ultrasonography and Doppler in the management of growth restricted fetuses is reported in the Growth Retardation Intervention Trial (GRIT) study22. This study concluded that crude umbilical Doppler assessment together with obstetric uncertainty and the option of either immediate delivery or further assessment could not change perinatal mortality. These findings have indirectly strengthened the non evidence-based but strong conclusion that the management of severe intrauterine growth restriction largely relies on biometric and intrafetal Doppler assessment through monitoring of adaptative changes in fetal hemodynamics under hypoxia and acute alteration under critical acidotic conditions23. This appears to be the best available management until a randomized study can eventually fine tune the optimal delivery conditions of a compromised fetus24.
The routine use of Doppler ultrasonography in pregnancy is not advisable according to RCTs20, 25. However, uterine artery Doppler in the mid-trimester has a sensitivity of about 60% for detecting severe cases of fetal growth restriction and pre-eclampsia, at the cost of following-up 5% of the general population12. The ‘missing link’ between diagnosis and favorable outcome is the lack of a suitable pharmaceutical or obstetric solution.
Although the majority of cases with shoulder dystocia and brachial plexus injury are unpredictable and occur in non-macrosomic neonates, the risk is increased with macrosomia, especially so in the context of diabetes26. However, in non-diabetic pregnancies planned interventions based on estimates of fetal weight do not reduce the incidence of shoulder dystocia and do not decrease adverse outcomes attributable to fetal macrosomia27. Detection of a macrosomic fetus in an uncomplicated pregnancy ranges from 15 to 79%, with sonographic estimates of birth weight, and from 40 to 52% with clinical estimates. Among diabetic patients the probability of identifying a newborn weighing > 4000 g clinically and sonographically is over 60%28. Based on an 87% accuracy of detecting macrosomia, only 10.6% of diabetic patients at term require intervention (6.8% labor induction, 3.8% elective Cesarean section) and the rate of shoulder dystocia is 7.4% in macrosomic infants delivered vaginally26. Ultrasonography has therefore become pivotal in the management of diabetes in pregnancy.
Placenta previa occurs in 0.5–1% of all pregnancies, with ultrasonography the best method available for placental assessment. At 20–23 weeks' gestation a cut-off of 25 mm between the placental edge and the internal os predicts placenta previa at delivery with a false positive rate of < 1%, defining, by ultrasonography, a high risk of maternal and perinatal severe morbidity and mortality29. An ultrasound-based intervention reduces the overall Cesarean section rate from 90 to 30% as well as the overall need for blood transfusion30. Vasa previa is a rare event associated with a high perinatal mortality. Its prenatal ultrasound diagnosis is reliable and may increase survival from 44 to 94% in cases subsequently managed by elective Cesarean section at 35 weeks' gestation31.
Ultrasound assessment of the uterine scar of a previous Cesarean section by measuring the thickness of the lower uterine segment before labor has an excellent negative predictive value (99.3–100%) for the risk of uterine defect. This is likely to support the development of an individual assessment, in combination with maternal characteristics and previous history, of the risk of uterine rupture during labor32. In up to 93% of the cases from this population, all efforts made at identifying placenta accreta by ultrasonography and magnetic resonance imaging (MRI)33 may be further justified if conservative management at delivery can decrease maternal morbidity as compared with extirpative management34.
Heads or tales?
The contribution of ultrasonography to the assessment of the cervix and the fetal head position prior to the induction or spontaneous onset of labor is well recognized35–38, but controversial39–42, depending on the characteristics of the population (e.g. the proportion of unripe cervices, gestational age and of indications for induction, i.e. post-term or high-risk cases) and the criteria chosen (success rate of induction; likelihood of a vaginal or Cesarean section delivery39, 43, 44 or Cesarean section with dilatation < 8 cm42; induction to delivery43; latent-phase38, 40, 41 or overall36, 37 duration of labor; and even vaginal delivery within 24 h35, 39). Although all these studies were performed in cases with a Bishop score of less than 5, the best results are reported in post-term pregnancies with a ripe cervix in up to 60% of cases44 and using a cut-off of < 26 mm for cervical length42. In this situation again, the apparent uncertainty about ultrasonography's abilities cannot be attributed to the technique but mainly to the absence of standardization of the parameters used to assess the cervix both clinically (by Bishop score or one of its components) and ultrasonographically (by cervical length, dilatation, presence of wedging and cervical angle) as well as the method of induction. Only two studies provide sample size calculations40, 41 involving more than 100 patients. Secondary exclusion of up to 41% of cases, all with Cesarean deliveries35, also impacts on the conclusions drawn. Ultrasound measurement of the cervical length at between 39 + 4 and 40 + 3 weeks of gestation was predictive of the onset of spontaneous labor within 7 days, while fetal fibronectin assay was useful to predict the risk of Cesarean delivery45. It was hypothesized that ultrasound examination of the cervix and fetal fibronectin assessment addressed two different physiological processes involved in parturition. In predicting the time to the onset of spontaneous labor, cervical length may be an indicator of myometrial ripeness but it does not appear, unlike fetal fibronectin, to be one of cervical ripening45–47. The absence of myometrial ripeness among patients requiring prostaglandins for induction of labor may explain the lack of correlation between cervical length and the total duration of labor.
Antepartum oligohydramnios, defined as a sonographically determined amniotic fluid index of ⩽ 50 mm, is associated with increased perinatal morbidity and mortality and Cesarean delivery for fetal distress48, 49. However, amniotic fluid index seems to overestimate the number of abnormal results in post-term pregnancies and therefore may cause, rather than prevent, morbidity, and the deepest pool should be considered instead50.
Intrapartum transabdominal ultrasonography increases the accuracy of the assessment of fetal head position during all stages of labor, whereas digital examination is wrong in 46–76% of cases even in experienced hands, and especially in occiput–posterior and occiput–lateral positions51, 52. Transabdominal ultrasound examination can also reliably diagnose head engagement and can help plan operative delivery53, 54.
Down to the floor
In the postpartum period, ultrasound examination of the anal sphincter improves detection of postpartum defects in up to 35–40% of all deliveries prior to the development of clinical symptoms, and confirms the lesion in most symptomatic women. Its routine use is gaining wide acceptance and may play a major role in the near future in the choice of the mode of delivery in subsequent pregnancies. This may be done together with the assessment of two other risk factors amenable to ultrasound examination: macrosomia and occipito–posterior position of the fetal head55.
Ultrasound-driven evidence-based obstetrics
Most general obstetricians do not perform ultrasound themselves, which implies that ultrasonography is likely to remain underused56. Moreover the problem of the quality of the ultrasound examination arises early since only 89.4% correctly assess the gestational age of the pregnancy (±1 week)57 and only approximately one fifth of obstetrics and gynecology residents opt to use this diagnostic modality in clinical practice58. There is no such thing as a ‘quick scan’ even in targeted examinations and it is time for residency programs to give obstetric ultrasonography some well-deserved consideration. The time has come to move away from a paternalistic approach of determining what amount of ultrasound assessment can be offered to high- or low-risk pregnant women and to rely on an individual risk assessment for a given false positive rate (usually set at 5% at 22–24 weeks' gestation). The model established for ultrasound-based fetal aneuploidy screening can be applied to construct risk assessment algorithms for pre-eclampsia, prematurity and uterine rupture after previous Cesarean section as well as growth restriction and macrosomia. Only then would we have ultrasound-driven evidence-based obstetrics.