Estimation of Fetal Weight

Accuracy of obstetric ultrasonography for estimating fetal weight

Obstetric sonographic assessment for the purpose of obtaining fetal biometric measurements to predict fetal weight has been integrated into the mainstream of obstetric practice during the past quarter century. From its inception, this method has been presumed to be more accurate than clinical methods for estimating fetal weight. The reasons for this assumption are varied, but the fundamental underlying presumption is that the sonographic measurements of multiple linear and planar dimensions of the fetus provide sufficient parametric information to allow for accurate algorithmic reconstruction of the 3-dimensional fetal volume of varying tissue density. Consistent with these beliefs, much effort has generated best-fit fetal biometric algorithms that can make birth weight predictions based on obstetric ultrasonographic measurements. As such, the ultrasonographic technique represents the newest and most technologically sophisticated method of obtaining birth weight estimations.

Modern algorithms that incorporate standardly defined fetal measurements (eg, some combination of abdominal circumference [AC], femur length [FL], and either biparietal diameter [BPD] or head circumference [HC]) are generally comparable in terms of overall accuracy in predicting birth weight. The most commonly used fetal biometric algorithms are shown in Table 6. When other sonographic fetal measurements are used to estimate fetal weight (eg, humeral soft tissue thickness, subcutaneous tissue/FL ratio, cheek-to-cheek diameter), these nonstandard measurements do not improve significantly the ability of obstetric sonography to predict birth weight, except in special patient subgroups (eg, mothers with diabetes).

Table 6. Ultrasonographic Fetal Biometric Prediction Algorithms for Calculating Estimated Fetal Weight*

Ln - Natural logarithm

BW - Estimated fetal weight (g)

BPD - Fetal biparietal diameter (mm)

HC - Fetal head circumference (mm)

AC - Fetal abdominal circumference (mm)

FL - Fetal femur length (mm)

In a recent study of 1034 patients, the mean absolute percentage error associated with the calculation of estimated fetal weights based on fetal measurements of BPD, AC, and FL (according to a widely used equation of Hadlock) was 10.0-11.3%, depending on the gestational age of the fetus (ie, after a crude stratification of fetal size). When the mean absolute percentage error of the method is assessed for 3 different clinically significant ranges of fetal weight (ie, <2500 g, 2500-4000 g, >4000 g), the mean absolute percentage error of the technique typically is lowest (?.1-10.5%) for the mid range (2500-4000 g) and higher values of fetal weight (>4000 g) and slightly greater for fetuses weighing less than 2500 grams (?.0-11.0%). When another commonly used measure of accuracy is used (the percentage of fetuses with weight accurately estimated to within ?0% of actual birth weight), 56% were predicted accurately to within these limits for fetuses weighing less than 2500 grams, 58% for fetuses weighing2500-4000 grams, and 62% for fetuses with actual birth weights greater than 4000 grams.

When the accuracy of the detection of clinically relevant deviations in term birth weight is assessed using the sonographic technique (ie, ability of the sonographic method to accurately identify term fetuses weighing <2500 g, >4000 g, and >4500 g), the positive predictive value is 44-55%, with associated sensitivities of 58-71%. For preterm fetuses delivered at less than 37 weeks' gestation, the 1-way accuracy of such sonographic fetal biometric classifications of clinically significant birth weight deviations (ie, low birth weight) is better; the positive predictive value of a sonographic estimate of fetal weight less than 2500 grams is 87% for preterm fetuses, with an associated sensitivity of 90%, and the positive predictive value for a sonographic estimate of fetal weight less than 1500 grams is 86%, with an associated sensitivity of 93%.

The overall results for the sensitivity, specificity, positive predictive value, and negative predictive value of the sonographic technique for predicting significant variations in fetal weight are shown as a function of both fetal weight and gestational age in Table 7.

Results for 5 different studies that investigated the accuracy of the technique for predicting fetal macrosomia of greater than 4000 grams at term are shown in Table 8. When a meta-analysis of these 5 studies incorporating 2367 term pregnancies of greater than 37 completed weeks' gestation was performed, the positive predictive value for detecting fetal weight greater than 4000 grams using the sonographic fetal biometric technique was 59% and the associated sensitivity was 59%. The average predictive error in birth weight estimates for fetuses of greater than 4000 grams using this method was routinely greater than 300-400 grams.

Table 7. Accuracy of Sonographic Fetal Biometry for Detecting Clinically Relevant Deviations of Fetal Weight*

* Adapted from Chauhan SP et al
Table 8. Accuracy of Obstetric Ultrasonography for the Prediction of Fetal Macrosomia of >4000 g at >37 Weeks' Gestation

The notion that multiple obstetric sonographic fetal biometric evaluations might prove superior to a single examination for predicting fetal weights has been examined. One recent study evaluated the advantage of multiple ultrasonographic examinations compared with a single examination for the purpose of estimating fetal weight. The accuracy of birth weight percentile predictions was similar whether 1 or multiple such examinations was performed during the third trimester. In this study, which used the ultrasonic algorithm of Shepard, 38% of the fetuses had their weight accurately estimated to within ?0% after a single ultrasonographic assessment of fetal dimensions, and 42% had such predictions correct to within ?0% after multiple sonographic examinations were performed. No statistically significant difference occurred in accuracy between these 2 approaches.

The sensitivity, specificity, positive predictive value, and negative predictive value for the prediction of both small-for-gestational-age and large-for-gestational-age fetuses using these sonographically derived estimated fetal weights, which are obtained from 1 or more sonographic examinations, are shown in Table 9.

Table 9. Accuracy of Single vs Multiple Sonographic Fetal Biometric Examinations for Detecting Clinically Relevant Deviations in Fetal Weight*

* Adapted from Hedriana HL et al

^{?/SUP> Prevalence of SGA fetuses in the series was 7.2% (19 of 264 patients), and the prevalence of LGA fetuses was 17.4% (46 of 264 patients).}

Another question is the potential difference in the predictive accuracy of fetal weight estimates made using fetal biometric measurements obtained by professional sonographers in a controlled setting compared with hospital-based resident physicians performing studies in a labor and delivery unit. Although the interobserver variation in ultrasonic fetal biometric measurements has been shown to be small, these differences still may introduce unacceptable variability into the parameters employed for fetal weight estimation by fetal biometric algorithms.

In a recent study designed to address this clinically important question, the mean absolute percentage error associated with ultrasonographic estimates of fetal weight by house staff physicians in a labor and delivery suite (?.3%) was comparable to that reported by professional ultrasonographers in a controlled setting. Thus, no clinically important systematic bias is introduced into such results based on differences in operator training or diagnostic setting.

Several technical limitations of the sonographic technique for estimating fetal weight are well known. Among these are maternal obesity, anterior placentation, and oligohydramnios.

Recently, several studies challenged the overall accuracy of sonographic birth weight estimations. More than a dozen investigations concluded that ultrasonography may be no more accurate for predicting birth weight than clinical palpation or even maternal self-estimations of fetal weight. Two of these studies also suggested that quantitative assessment of maternal characteristics may be as accurate as obstetric ultrasonography for the purpose of predicting the occurrence of fetal macrosomia.

Maternal self-estimations of fetal weight

Recently, 3 studies examined the accuracy of patient self-estimations of fetal weight by parous women. The mean absolute percentage errors for these birth weight predictions was 8.7-9.5% for term fetuses, with mean absolute birth weight errors of 305-350 grams. In a small study that reported the sensitivity for macrosomia greater than 4000 grams, it was 56% (see Table 10). These results seem comparable to those reported for both clinical palpation and obstetric ultrasonography.

Table 10. Comparison of Published Accuracy of Different Methods for the Prediction of Term Fetal Macrosomia >4000 g

Recently, a new, theoretically defensible equation that can predict individual birth weights prospectively from maternal characteristics was developed. To do this, the predictive efficacy of 59 scientifically justifiable terms was evaluated simultaneously, obviating any confounding covariation and determining which of the predictors could account for variations in birth weight that others could not. Aside from maternal race, only 6 maternal and pregnancy-specific variables were important in the prediction of birth weight for otherwise normal gravidas. Only 1 additional paternal factor was found to be independently predictive of birth weight (ie, paternal height), but it accounts separately for less than 2% of the variance. The first-order correlation of each of these predictors of fetal weight is shown in ^{Table 3.}

Using these routinely recorded variables, an equation based on maternal demographic and pregnancy-related characteristics alone was developed (Equation 1) to predict birth weight based on the following:

• Maternal height
• Maternal weight at 26 weeks' gestation
• Maternal weight gain rate during the third trimester
• Parity
• Fetal gender
• Gestational age at delivery

These prospectively measurable variables can explain 36% of the variance in term birth weight and can predict birth weight accurately to within ?67 grams (?.6% of individual birth weights). In addition, 75% of newborn weights can be estimated properly to within ?0% of actual birth weight using this technique. The equation generated for this purpose is as follows:

Equation 1

Birth weight (g) = gestational age (d) x [9.36 + 0.262 x fetal gender + 0.000237 x mat. height (cm) x mat. weight at 26.0 wk (kg) + 4.81 x mat. weight gain rate (kg/d) x (parity + 1)]

Where:

Fetal gender = +1 for males, -1 for females, and 0 for unknown gender

Gestational age = days since onset of last normal menses = conception age (d) + 14

The findings are as follows:

• The trade off between sensitivity and specificity for the prediction of fetal macrosomia greater than 4000 grams is shown in Picture 1.

  　

• The trade off between the positive and negative predictive value for the prediction of birth weight greater than 4000 grams is shown in Picture 2.

  　

• The overall accuracy of the equation for the prediction of fetal macrosomia is shown in a receiver operating characteristic curve in Picture 3. The area under this curve is 0.82.

  　

• Using a cutoff value of 3775 grams, the sensitivity for the prediction of fetal macrosomia greater than 4000 grams is 54%, the specificity is 92%, the positive predictive value is 52%, and the negative predictive value is 93%.

These results are consistent with those reported previously using a similar quantitative maternal characteristics approach.

Which of the methods for predicting fetal weight is the most accurate?

The accuracy of the different methods of predicting fetal weight depends on the gestational age and range of birth weights under study. Again, for this purpose, dividing fetuses into 3 birth weight categories of less than 2500 grams, 2500-4000 grams, and greater than 4000 grams is useful. The relative accuracy of clinical palpation versus obstetric sonographic fetal biometry for these 3 birth weight ranges is shown in Table 11 (see below). For the clinically significant birth weight ranges of less than 2500 grams and greater than 4000 grams, the accuracy of sonographic fetal biometry appears to be superior to clinical palpation for predicting the occurrence of low birth weight fetuses weighing less than 2500 grams, whereas the 2 techniques appear to be comparable in predictive accuracy for fetuses weighing 2500 grams or greater.

Table 11. Accuracy of Clinical Palpation vs Sonographic Fetal Biometry for Predicting Actual Birth Weight <2500 g, 2500-4000 g, and >4000 g*

Abs % error - Mean absolute percentage error of fetal weight predictions (%)

% within 10% - % of fetuses with weight accurately predicted to within 10% of actual birth weight

* Adapted from Chauhan SP et al and Sherman DJ et al

A recent study directly comparing the 4 different methods of fetal weight prediction in 44 normal term pregnancies found that there was no difference between the accuracy of the clinical methods (eg, clinical palpation, birth weight prediction equation, maternal self-estimation of fetal weight) and ultrasonic fetal biometric techniques for predicting term birth weight. Eight different ultrasonic fetal biometric algorithms were assessed for this purpose. The mean birth weight for newborns in this study was 3445 ?58 grams, with a birth weight range of 2485-4790 grams. These results are summarized in Table 12 (see below). No systematic advantage was found with the ultrasonic technique for predicting term birth weight over the clinical methods.

Seven other recent studies directly compared the accuracy of clinical palpation to ultrasonographic fetal biometry using the same gravidas (see Table 5), and 3 compared clinical palpation to parous patients' self-estimates of fetal weight after 37 completed weeks' gestation.

One study compared clinical palpation to both ultrasonographic fetal biometry and parous patients' self-estimations of fetal weight. All of the methods have significant predictive errors in birth weight estimations for term fetuses that range from 290-560 grams, and no consistent or clear superiority of ultrasonographic fetal biometry over the other techniques of fetal weight estimation was found.

Table 12. Comparison of Results for Different Methods of Predicting Term Birth Weight*

* Study of 44 term patients by the author
　

DIAGNOSING SIGNIFICANT DEVIATIONS IN FETAL WEIGHT AND MANAGEMENT OPTIONS

Developing a consensus of indicators

All currently available techniques for estimating fetal weight have significant degrees of inaccuracy. Wikstrom et al demonstrated that by combining clinical and ultrasonographic data about fetal size, an improved accuracy in fetal weight estimations can be obtained. Based on this finding, a reasonable strategy for arriving at estimated fetal weight is to utilize multiple estimates based on different sources of clinical and sonographic information. If such a strategy is accepted, then a practical and semiquantitative schema for making an accurate antenatal diagnosis of fetal weight in the clinical setting can be suggested, as follows:

　

• First, assess maternal risk factors for predispositions to fetal growth deviations at the initial prenatal visit and again at the start of the third trimester. Any gravida who has 1 or more of the following conditions should be considered at high risk for abnormal fetal growth and should undergo further assessment via other techniques to estimate their fetal weight:

  　

  • Poorly controlled diabetes mellitus (any class, including gestational diabetes)
  • Abnormal 1-hour glucose screening test greater than 135 mg/dL
  • Single abnormal value on 3-hour oral glucose tolerance testing
  • Obesity
  • Abnormally tall or short stature
  • Excessive or inadequate pregnancy weight gain
  • High parity
  • Preterm gestation
  • Postdate pregnancy
  • Chronic hypertension
  • Preeclampsia (including pregnancy-induced hypertension, HELLP syndrome)
  • Microvascular disease
  • Cigarette smoking
  • Residence at high altitude
• Once a gravida is thought to be at risk for either excessive fetal growth or intrauterine growth restriction, employ all applicable test modalities to determine which of the different methods (if any) suggest that the fetus has a weight outside of the reference range for its gestational age.
  • The standard fetal weight for comparison can be obtained from the 50th percentile rank of published fetal growth curves that are derived from patient populations that are well matched to the particular patient under consideration.
  • Clinical palpation and ultrasonographic fetal biometry can always be used to obtain an estimate of fetal size, regardless of gestational age.
  • Additionally, the current version of the birth weight prediction equation can be used to estimate the upper fetal weight limit for pregnancies in normal gravidas of all races at or near term.
  • If any 2 modalities suggest that the fetal weight is abnormal for gestational age (ie, >2 SD from the expected mean or 50th percentile rank value), then presume that the fetus is growing at an abnormal rate.
  • If so, then institute serial assessments of fetal growth to determine the velocity of ongoing fetal weight gain.
  • If both fetal weight for gestational age and the velocity of fetal weight gain are abnormal, the evidence for a significant abnormality in fetal growth becomes more compelling. Under such circumstances, manage the pregnancy accordingly, with the presumption that the fetal size is outside the reference range of expected values.

Option for suppression of labor in women carrying undersized fetuses

In general, the case can be made to attempt labor suppression in women carrying preterm fetuses weighing less than 2000-2500 grams. As stressed previously, most low weight fetuses are associated with preterm gestations. However, any recommendation in this circumstance regarding tocolysis presupposes the following: (1) no immediate fetal or maternal indications mitigate toward the timely delivery of the undersized fetus, and (2) the undersized fetus will continue to grow along an acceptable growth curve if the gestation is allowed to continue. In many cases, both of these assumptions are invalid. For instance, many women who are delivered of preterm neonates are allowed to do so because of compelling fetal or maternal medical conditions that warrant timely delivery (eg, intrauterine infection, severe uteroplacental insufficiency, severe preeclampsia). If fetal infection or IUGR are present, the preterm delivery of an underweight fetus may be indicated.

The increased risk of perinatal complications associated with the delivery of an underweight fetus in these circumstances may be outweighed entirely by the increased risk of morbidity and mortality of allowing the pregnancy to continue for both the fetus and mother. Additionally, in some circumstances, the inadequate velocity of fetal growth might mandate a decision for delivery. In such cases, the presumption is that extrauterine growth and development in the neonatal nursery would be superior to that achieved in utero. Clinical judgment under such circumstances is of paramount importance in deciding when to effect delivery and when to attempt labor suppression. More detailed considerations for aiding in this decision are beyond the scope of this chapter.

Option for labor induction in women carrying oversized fetuses

For fetuses delivered before 37 weeks' gestation, fetal macrosomia is a rarity; more than 99% of macrosomic fetuses are the product of term gestations. In general, nearly 95% of fetuses will gain 12.7 ?.8 grams per day from 37-42 weeks' gestation, meaning that an average fetus gains an additional 445 ?8 grams (1 lb ? oz) during this period. If a patient is thought to have a term fetus weighing more than 4000 grams and is willing to undergo labor induction, effecting vaginal delivery in these gravidas sooner, rather than awaiting the onset of spontaneous labor and a higher average birth weight at delivery, is often reasonable.

In studies that have attempted to examine this question, labor induction has not been demonstrated conclusively to decrease the fetal and maternal risks of intrapartum complications, and the cesarean delivery rate has been suggested to increase in several studies, whereas it has been purported to be unchanged in others. The difficulty in interpreting these results is that there have been significant differences among the predicted and actual birth weights for patients included for investigation and the power of the studies so far conducted have been insufficient to conclusively demonstrate statistically significant differences in adverse fetal outcomes among different study groups.

As with the case of preterm delivery of underweight fetuses, many considerations, including the size of the maternal pelvis and the weight of previously delivered fetuses, should be taken into consideration. Clinical judgment in these circumstances is of paramount importance in deciding whether or not labor induction is indicated in an attempt to minimize excessive fetal weight at delivery.

Conclusions

Both low birth weight (<2500 g) and high birth weight (>4000 g) are fetal conditions that are associated with increased risks of peripartum morbidity and mortality. Although the absolute risk that fetuses with birth weights of 2000-2500 grams and 4000-4500 grams will suffer major peripartum complications is not overwhelming, the risk of such complications increases substantially with both decreasing and increasing birth weight relative to these lower and upper limits. Thus, birth weight and gestational age are both important determinants of peripartum outcome. From this standpoint, the optimal range of newborn weight generally is thought to be 3000-4000 grams (6 lb 10 oz to 8 lb 13 oz). As always, the problem is knowing the fetal weight with sufficient accuracy in advance of delivery.

Many factors that impact directly upon birth weight are not modifiable. These include maternal race, height, parity, paternal height, and fetal gender. However, what can be influenced with potentially significant effects upon birth weight are the following:

　

• Prepregnancy weight
• Pregnancy weight gain
• Glucose control in patients with diabetes or glucose intolerance of pregnancy
• Gestational age at delivery

All of these factors can have significant impacts on fetal weight at delivery. Whereas permitting the delivery of fetuses that weigh 2000-2499 grams typically is not associated with an overwhelming increase in neonatal complications compared with normal weight neonates, those fetuses weighing less than 2000 grams at birth are at increased risk for perinatal complications in a manner that is commensurate with their weight.

Similarly, whereas allowing a trial of a vaginal delivery for a fetus estimated to weigh 4000-4499 grams may be reasonable in many circumstances, many sources suggest that fetuses with estimated weights of 4500 grams or greater should be delivered by cesarean birth, in order to avoid the increased intrapartum risks associated with the vaginal delivery of a macrosomic fetus. This is especially true when gestational diabetes is involved and the fetal conformation may be altered to reflect a larger shoulder girdle or head circumference ratio compared with the offspring of mothers without diabetes.

In the case of macrosomic fetuses, attempts to predict birth weight from fetal measurements obtained via ultrasonography have proven unsuccessful from the standpoint of improving clinical outcomes. Many studies conclude that ultrasonographic fetal biometric assessments are no more predictive of fetal macrosomia than clinical assessments of fetal size by simple external abdominal palpation (see Table 5 and Table 10). Both ultrasonography and manual assessment of fetal size have sensitivities of less than 60% for the prediction of fetal macrosomia, with false positivity rates greater than 40%. Likewise, for small fetuses less than 1800 grams, ultrasonic fetal weight estimates often are in error by as much as 25%.

By using a birth weight prediction equation that is based on maternal and pregnancy-specific characteristics alone, fetal weight at and near term can be predicted with a high degree of accuracy (?.6%). This approach appears to be at least as reliable for predicting fetal macrosomia in healthy gravidas as both clinical palpation and ultrasonographic fetal biometry, neither of which can be used with any degree of certainty in advance of the date of delivery. Such a quantitative assessment of maternal characteristics serves to objectively quantify the majority of previously recognized clinical variables that have long been employed in subjective clinical assessments and that are thought to be predictive of fetal weight.

By contrast, clinical palpation is a subjective methodology that must be employed at or near the date of delivery, and it is both patient and clinician-dependent for its success (ie, less accurate for obese than nonobese gravidas, significant for interobserver variation in birth weight predictions even among experienced clinicians).

The disadvantages of ultrasonographic fetal biometry are that the method is both complicated and labor-intensive, potentially being limited by suboptimal visualization of fetal structures. It also requires costly sonographic equipment and specially trained personnel. Although such expensive imaging equipment is widely available in the US and other industrialized countries, this is generally not the case in developing nations where a scarcity of medical resources often occurs.

In the future, combining the different methods of fetal weight prediction to improve their overall accuracy may be possible. Wikstrom et al suggested that by combining the independent information about fetal size obtained from the 3 different approaches (ie, clinical examination, quantitative assessment of maternal characteristics, ultrasonographic fetal biometry) the predictive value of fetal weight estimations can be improved dramatically. In the case of excessive fetal size, combining these methodologies may result in an 80% positive predictive value for the identification of fetal macrosomia with a sensitivity of 63% and specificity of 95%.

Recently, a quantitative combination of maternal demographic information (of the type incorporated in Equation 1) with the independent information obtained by ultrasonic fetal biometry (abdominal circumference) has been demonstrated to improve birth weight prediction substantially, with the area under the ROC curve increasing to 0.92. The mean absolute percentage error in birth weight predictions that can be attained using this new combinatorial method is ?.4%.

With the advent of 3-dimensional fetal imaging, optimism that these new technologies can provide even better fetal weight estimations may be justified, but the advantages of estimating fetal weight using these newer techniques have not yet been demonstrated. Using these new approaches, further improvements in the accuracy of fetal weight prediction in the future will permit prospective obstetric intervention to be undertaken more confidently by practicing obstetricians, with the aim of minimizing intrapartum and peripartum risks for both fetuses and mothers.

PICTURES

Caption: Picture 1. Sensitivity and specificity for the prediction of fetal macrosomia >4000 g using Equation 1. Jagged lines represent results obtained from raw data; curved lines represent smoothed values.

Picture Type: Graph

Caption: Picture 2. Curve delineating the trade-off between positive and negative predictive value using Equation 1

Picture Type: Graph

Caption: Picture 3. Receiver operating characteristics curve for the prediction of fetal weight >4000 g using Equation 1. The dashed lines represent the sensitivity and specificity associated with prediction equation cutoff values of 3500 g, 3700 g, and 3920 g, respectively.

Picture Type: Graph