| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
by the American Institute of Ultrasound in Medicine J Ultrasound Med 22:347-356 • 0278-4297 Adverse Birth Outcomes in Relation to Prenatal Sonographic Measurements of Fetal SizeDepartments of Radiology (R.S.-B., P.W.C., V.A.F., R.A.F.) and Epidemiology and Biostatistics (R.S.-B., P.B.), University of California, San Francisco, California; and Department of Obstetrics and Gynecology (J.E.), Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts USA. Address correspondence and reprint requests to Rebecca Smith-Bindman, MD, Department of Radiology, University of California San Francisco, UCSF/Mount Zion Campus, 1600 Divisadero St, San Francisco, CA 94115 USA.
Objective. To evaluate and quantify the prediction of multiple neonatal outcomes by sonographically estimated fetal weight across a broad range of gestational ages. Methods. A retrospective cohort analysis was conducted among women with certain gestational age (n = 1376) seen at the University of California San Francisco from 1994 through 1997. The relative risks for small size at birth, small (low birth weight) for gestational age, and adverse neonatal outcomes were compared between small and average-sized fetuses. Results. Fetuses with an estimated fetal weight in the 5th percentile or less for gestational age were at increased risk of a birth weight less than 2000 g (relative risk, 6.5), a birth weight in less than the 3rd percentile for gestational age (relative risk, 10.1), preterm birth (relative risk, 2.2), extreme preterm birth (relative risk, 5.7), prolonged neonatal hospital stay (relative risk, 2.7), neonatal intensive care unit admission (relative risk, 3.2), and stillbirth or neonatal death (relative risk, 7.7) compared with average-sized fetuses (all P < .0001). With intrauterine growth restriction defined as an estimated fetal weight in the 5th percentile or less for gestational age, up to 29% of fetuses with adverse neonatal outcomes were detected, for false-positive rates of only 4% to 5%. After adjusting for confounding variables, low estimated fetal weight remained a significant predictor of neonatal morbidity and mortality. Conclusions. Morbidity and mortality are significantly increased among fetuses with an estimated fetal weight in the 5th percentile or less for gestational age.
Key Words: birth outcomes • fetal biometry • fetal weight • intrauterine growth restriction • neonatal morbidity • neonatal mortality • neonatal outcomes • prenatal sonography Abbreviations: AC, abdominal circumference • BPD, biparietal diameter • EFW, estimated fetal weight • FL, femur length • GA, gestational age • HC, head circumference • IUGR, intrauterine growth restriction • LBW, low birth weight • LMP, last menstrual period • NICU, neonatal intensive care unit • SGA, small for gestational age • UCSF, University of California San Francisco
Infants who are small for gestational age (SGA) at birth (low birth weight [LBW]) have higher neonatal morbidity and mortality and worse long-term prognoses than infants who are appropriate size for gestational age (GA).1,2 In an attempt to identify fetuses at greatest risk for adverse neonatal outcomes, sonography has been used to identify small size in utero at varying points in gestation. Because it is often difficult to distinguish constitutionally small fetuses from fetuses whose sizes are altered by a pathologic process (such as placental insufficiency), all small fetuses are usually classified together and considered to have intrauterine growth restriction (IUGR).3,4 Although many definitions of IUGR have been used, the most common definition is an estimated fetal weight (EFW) in less than the 10th percentile for GA at any point in gestation.4 Sonographic measurements of the fetal head, abdomen, and extremities are used to estimate weight,5 which is compared with a population-derived distribution of fetal weight at the same GA. Fetuses with weight below a certain percentile are defined as having IUGR. Hence, size, rather than growth, is most often used to define IUGR,6 although other definitions of IUGR have been suggested.7–9 As a routine part of obstetric care, sonography is widely used to assess IUGR in the second and third trimesters of pregnancy. However, few studies have actually quantified the association between sonographically defined IUGR and neonatal morbidity and mortality.10 A number of studies have described the association of EFW immediately before delivery with birth weight, suggesting that sonography is a fairly accurate method of estimating fetal weight late in pregnancy.11–15 Other studies have evaluated the association of IUGR diagnosed on the basis of sonography and LBW but not other adverse neonatal outcomes.16–18 However, most of these reports were case-control studies, making it difficult to evaluate the predictive value of sonography. One study identified an association between sonographic EFW in the first trimester of pregnancy and birth weight, birth weight for GA, and prematurity.19 However, in clinical practice, most women do not have ultrasound examinations until the second or third trimester of pregnancy. We sought to determine and quantify which aspects of neonatal morbidity and mortality could be predicted using second- and third-trimester sonographic measurements of fetal size. We also sought to ascertain which sonographic measurements of size best identify those fetuses at greatest risk of neonatal morbidity and mortality.
Study Population Our population included 1836 women with singleton pregnancies who underwent at least 1 second- or third-trimester obstetric ultrasound examination in the Department of Radiology at the University of California San Francisco (UCSF) Medical Center at 13 to 37 weeks’ gestation from July 1994 through March 1997 and who subsequently gave birth at the UCSF Medical Center. We excluded examinations performed only to assess amniotic fluid volume, cervical length, or biophysical profile. We also excluded women with 5 or more ultrasound examinations during a single pregnancy, multiple pregnancies reduced to singleton pregnancies, women who had undergone fetal surgery, and women transferred to UCSF for delivery, as well as fetuses with major congenital or chromosomal anomalies (noted in the UCSF ultrasound database), because such fetuses are more likely to be small and to have neonatal morbidity unrelated to their size. Among these 1836 women, a subset of 236 had 2 or more ultrasound examinations, with which we evaluated growth (change in weight over time), as described elsewhere.9 The UCSF Institutional Review Board approved the study, and a waiver for informed consent was obtained.
Sources of Data The UCSF ultrasound database was linked with the UCSF obstetric database, which includes demographic information and details of obstetric histories and neonatal outcomes for all births at UCSF. Variables abstracted from the obstetric database included maternal race or ethnicity, prepregnancy height and weight, history of substance abuse, neonatal birth weight, sex, GA at delivery, length of neonatal hospital stay, neonatal intensive care unit (NICU) admission, length of NICU stay, assisted ventilation at birth, stillbirth or neonatal death, and the presence of fetal anomalies. Neonatal outcomes are obtained in this database through linkage with the UCSF neonatal database. A senior obstetrician assigned GA on the basis of the patient’s records considering the following variables in descending order of importance: (1) in vitro fertilization date (assigned as the date of fertilization); (2) certain and regular LMP date consistent within 7 days of first-trimester sonography or within 14 days of second-trimester sonography; (3) first-trimester sonography; (4) second-trimester sonography20; and (5) third-trimester sonography. We compared estimated GA in the obstetric database with GA in the ultrasound database to ensure consistency. One of the authors reviewed inconsistent records, and 2 authors reached a consensus on how to date these pregnancies (<1%). We used the most accurate method of dating available in either database. Women in whom the GA was certain on the basis of in vitro fertilization dating, accurate LMP dating, or first-trimester sonography were considered to have certain GA (75%; n = 1376).
Calculation of Fetal Weight
Outcome Measures
Analysis We plotted the incidence of each outcome within each of the 7 EFW categories and generated receiver operating characteristic curves for each of the following definitions of abnormal sonographic findings: low EFW, small AC, small HC, small BPD, small FL, large HC/AC ratio, large BPD/AC ratio, and large FL/AC ratio. We performed multiple logistic regression to determine the association between EFW and adverse outcomes after adjusting for potential confounding variables, including maternal age (categorized as <21, 21–35, and >35 years), maternal height, maternal weight, body mass index, race or ethnicity (white, black, Hispanic, and other), parity (nulliparous versus multiparous), maternal substance abuse, and fetal sex.
In this racially and ethnically diverse sample, the rates of small size at birth, SGA, and adverse neonatal outcomes (Table 1  ) were very similar in women with certain and less certain GA (Table 1; all P > .05). The risks of adverse birth outcomes were higher than national averages. For example, 15.3% were born at a birth weights of less than 2500 g (compared with 6.4%–13.1% nationally depending on racial or ethnic group)23; 3.8% were born at less than the 3rd percentile for GA (compared with 3% nationally); and 18.3% were born preterm (compared with 8.9%–17.6% nationally depending on racial or ethnic group).23
As expected, fetuses with low EFW by sonography were at increased risk for LBW (Fig. 1A ), SGA (Fig. 1B), and adverse neonatal outcomes (Fig. 1C). In general, the lower the EFW percentile, the greater the relative risk of adverse outcomes. For example, a fetus with an EFW in the 5th percentile or less for GA was 3.0 times as likely to be born at a weight of less than 2500 g, 10.1 times as likely to be born at the 3rd percentile or less for GA, 2.2 times as likely to be born preterm, 5.7 times as likely to be born extremely preterm, 3.2 times as likely to be admitted to the NICU, and 7.7 times as likely to have stillbirth or neonatal death, compared with fetuses in the 21st-through-90th-percentile category (Table 2; all P < .0001). The risks of adverse neonatal outcomes were particularly raised for fetuses with EFW in the 5th percentile or less for GA (Fig. 1C). For instance, 28% of fetuses with EFW in the 5th percentile or less were born extremely preterm, compared with only 5% of average-sized fetuses. The relative risks for most adverse outcomes were also significantly increased for fetuses with EFW in the 20th percentile or less for GA (Table 3).
Although the 10th percentile has often been used to define abnormal sonographic findings, Figure 1C  shows lower observed rates of adverse outcomes in the 6th-through-10th-percentile group than in the 11th-through-20th-percentile group, probably reflecting stochastic variation due to small sample size but also suggesting that the 10th percentile might not be an optimal cutoff. The relationship between EFW and birth outcomes remained stable when FW was estimated in the earlier (13–24 weeks) and later (24–37 weeks) GAs.
Figure 2
Positive likelihood ratios are used to estimate the risk of adverse outcomes after a positive or abnormal test result, whereas negative likelihood ratios are used to estimate the risk of adverse outcomes after a negative or normal test result. A threshold at the 5th percentile of EFW was better at increasing the likelihood of small size at birth and adverse neonatal outcomes (positive likelihood ratios, 2.4–7.9) than normal sonographic findings were helpful in decreasing the risk of those same outcomes (negative likelihood ratio, 0.74–0.94; Table 4 ). Up to 37% of fetuses with EFW in the 5th percentile or less for GA were small at birth, were SGA, or had adverse neonatal outcomes (positive predictive values, 7%–37%). As would be expected, the positive predictive values were significantly and substantially lower when the 20th percentile cutoff was used to define an abnormal sonographic finding.
The individual biometric measurements were similar in their ability to detect fetuses who were small at birth, who were SGA, or who had adverse neonatal outcomes. The composite variable of EFW generally identified the greatest number of fetuses with abnormalities (highest sensitivity) at a given false-positive rate (Fig. 3
In multiple logistic regression, controlling for maternal age, race or ethnicity, height, weight, body mass index, substance abuse, prior preterm birth, and fetal sex, an EFW in the 5th percentile or less for GA remained a significant predictor of small size at birth, SGA, and adverse neonatal outcomes (adjusted odds ratios, 2.6–14.3; Table 5 ).
We found that fetuses with a low EFW, and particularly those in the 5th percentile or less for GA, were at increased risk of small size at birth, SGA, and adverse neonatal outcomes. This analysis suggests that a cutoff of EFW at the 5th percentile for GA, as opposed to the most widely used threshold of the 10th percentile, best identifies those fetuses at greatest risk. Compared with average-sized fetuses, fetuses with an EFW in the lowest 5 percentiles were 3 to 11 times more likely to be born with LBW, were 10 to 12 times as likely to be SGA, and had 2 to 8 times the risk of preterm birth, extreme preterm birth, long neonatal hospital stay, NICU admission, and stillbirth or neonatal death. An EFW in the 5th percentile or less for GA detected up to 29% of fetuses who were small at birth, were SGA, or had adverse neonatal outcomes, with false-positive rates of only 4% to 5% (Table 4 ). Additionally, among fetuses with EFW in the 5th percentile or less, 37% went on to have 1 or more adverse neonatal outcomes (Table 4). Additionally, although EFW is typically measured and reported only after 20 weeks’ GA,5 we found that EFW as early as 13 weeks’ GA was a stronger predictor of adverse birth outcomes. A birth weight in less than the 3rd percentile for GA is a particularly ominous outcome, because it strongly predicts neonatal morbidity and mortality for both full-term and preterm infants.2 The 5th percentile of EFW identified 24% of fetuses who later would be born at this very low-weight percentile (positive likelihood ratio, 5.4; negative likelihood ratio, 0.80). The 20th percentile of EFW detected 71% of the same fetuses, but at the expense of a much higher false-positive rate of 18% (positive likelihood ratio, 4.0; negative likelihood ratio, 0.35). Although it is not surprising that sonography facilitates prediction of adverse outcomes, little previous research has quantified the association between fetal size on second- and third-trimester prenatal ultrasound examinations and a broad range of neonatal outcomes. Understanding how sonography indicates such outcomes is essential for practitioners faced with the task of counseling patients after an ultrasound examination raises a question of small fetal size. The choice of a cutoff in EFW produces tradeoffs in predictive accuracy. Defining an abnormal sonographic EFW at the 5th percentile generates relatively high positive predictive values but low sensitivity, whereas a 20th percentile cutoff generates far better sensitivity but at the cost of many more false-positive diagnoses and lower positive predictive values. In the cohort studied, on the basis of the receiver operating characteristic curves generated, an EFW at or below either the 5th or 20th percentile for GA best identified those fetuses at greatest risk of adverse outcomes, and both thresholds performed better than the commonly used 10th percentile cutoff. We found little consistent difference in risk between the 6th and 20th percentiles. As a group, these fetuses had lower risks than the smallest fetuses but considerably higher risks than average-sized fetuses. The risk of adverse outcomes (and the associated positive predictive values) in our population was higher than US averages, in part reflecting the nature of our institution as both a community obstetric hospital and a tertiary referral hospital. Although the relationship between low EFW and adverse birth outcomes is likely to hold in other populations, the threshold that separates pregnancies with increased risk from those with average risk could differ. Incorporating race or ethnicity and including tests for interactions in all multivariate models failed to nullify the importance of size for predicting adverse outcomes, although the sample size was too small to determine whether normal fetal size differs by ethnicity or race. Additionally, we have not yet determined the best point in gestation to measure fetal size for predicting neonatal outcomes. Last, we included women with certain GA who had first-trimester sonography as the only method of dating, and this may have underestimated the importance of sonographically estimated size in predictive outcomes. The neonatal outcomes considered (including premature birth, requirement for ventilation at birth, NICU admission, and neonatal death) are complex and relate not only to maternal and placental factors but also to intrapartum complications. The fact that an ultrasound examination up to 25 weeks before delivery predicted as many as 29% of these outcomes is important. Although we found an association between the sonographic results and important clinical outcomes, we cannot be certain that the sonography led to decisions that caused, as opposed to predicted, some of the outcomes. For example, sonography may have led to the elective early delivery of an otherwise healthy neonate based on the suspicion of IUGR, and this would have artificially inflated the ability of sonography to predict outcomes. To reduce this potential bias, we excluded sonographic results after 37 weeks’ gestation. Furthermore, because sonography was able to predict such a broad range of outcomes, it seems likely that these outcomes were truly predicted, rather than iatrogenically caused, through the use of sonography. Additional research is needed to best determine how to manage and improve the outcomes associated with the accurate identification of growth-restricted (and not just constitutionally small) fetuses.24
Received October 1, 2002, from the Departments of Radiology (R.S.-B., P.W.C., V.A.F., R.A.F.) and Epidemiology and Biostatistics (R.S.-B., P.B.), University of California, San Francisco, California; and Department of Obstetrics and Gynecology (J.E.), Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts USA. Revision requested November 11, 2002. Revised manuscript accepted for publication December 12, 2002. We thank David Rock and Jeannie Rabold (Acuson; Siemens Medical Solutions, Mountain View, CA) for generous assistance in obtaining the data from the Aegis data system to complete these analyses and Travis Seawards for assistance in preparing the manuscript.
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Top
Abstract
Introduction
