abdominal circumference vs. estimated weight to predict large for gestational age birth weight in...

7
Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy William L. Holcomb Jr. a, *, Dorothea J. Mostello a , Diana L. Gray b a St. Louis University School of Medicine, 6420 Clayton Road, St. Louis, MO 63117, USA b Washington University School of Medicine, St. Louis, MO, USA Received 20 September 1999; accepted 1 March 2000 Abstract Early third trimester fetal abdominal circumference and sonographic fetal weight estimates were compared to predict large for gestational age birth weight in diabetic pregnancy. Both parameters have similar sensitivity, specificity, and predictive values. However, the optimal percentile cutoff values differ. Choice of birth weight standard significantly influences test characteristics. Negative prediction of large birth weight is more accurate than positive prediction. At third trimester sonography with maternal diabetes, the abdominal circumference percentile is potentially useful and should be routinely reported. D 2000 Elsevier Science Inc. All rights reserved. Keywords: Fetus; Ultrasound; Diabetes mellitus; Abdominal circumference; Large for gestational age 1. Introduction Almost 20 years ago, Ogata et al. [1] associated an abnormally large fetal abdominal circumference (AC) mea- sured between 28 and 32 weeks with ‘‘accelerated somatic growth’’ in diabetic pregnancy. Fetuses with AC measure- ments two standard deviations greater than the mean for gestational age became newborns with increased birth weight and increased skin-fold thickness. They also had increased amnionic fluid insulin levels. Seven years later, Bochner et al. [2] found that fetuses of diet-controlled gestational diabetic women with AC greater than the 90th percentile at 30–33 weeks had larger birth weight babies with increased shoulder dystocia, birth trauma, and Cesar- ean delivery for arrested labor. Subsequently, Buchanan et al. [3] performed a controlled clinical trial in which women with mild gestational diabetes and fetal AC greater than the 75th percentile at 29–33 weeks were randomly assigned to insulin + diet vs. diet alone. Women in the insulin group delivered babies with decreased skin-fold thickness who were less than one-third as likely to be large for gestational age (LGA) at birth. This demonstrated the potential utility of fetal sonographic information to select diabetic women for intensification of management. These findings raised several issues for us. Some built-in sonographic reporting systems do not routinely indicate a numerical AC percentile, but the estimated fetal weight (EFW) percentile is universally reported. Does the EFW in the early third trimester also predict LGA birth weight? There are various reference standards for fetal biometry and for birth weight as functions of gestational age. Are predic- tion characteristics similar with different reference stan- dards? Finally, does timing of the ultrasound examination within the early third trimester interval affect the relation- ship between AC and subsequent birth weight? The answers to these questions are not available in the literature to date. The objective of this study is to examine the test characteristics of early third trimester AC and EFW as predictors of LGA birth weight using different reference standards. 2. Materials and methods From mid-1993 to mid-1995, patients at a university- based specialty clinic for diabetic pregnancy routinely received fetal sonography between 28 0/7 weeks and 31 * Corresponding author. Department of Obstetrics and Gynecology, Division of Maternal– Fetal Medicine, St. Louis University, 6420 Clayton Road, St. Louis, MO 63117, USA. Tel.: +1-314-768-8873; fax: +1-314- 768-8776. 0899-7071/00/$ – see front matter D 2000 Elsevier Science Inc. All rights reserved. PII:S0899-7071(00)00153-4 Journal of Clinical Imaging 24 (2000) 1 – 7

Upload: william-l-holcomb-jr

Post on 17-Sep-2016

220 views

Category:

Documents


6 download

TRANSCRIPT

Page 1: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

Abdominal circumference vs. estimated weight to predict large for

gestational age birth weight in diabetic pregnancy

William L. Holcomb Jr.a,*, Dorothea J. Mostelloa, Diana L. Grayb

aSt. Louis University School of Medicine, 6420 Clayton Road, St. Louis, MO 63117, USAbWashington University School of Medicine, St. Louis, MO, USA

Received 20 September 1999; accepted 1 March 2000

Abstract

Early third trimester fetal abdominal circumference and sonographic fetal weight estimates were compared to predict large for gestational

age birth weight in diabetic pregnancy. Both parameters have similar sensitivity, specificity, and predictive values. However, the optimal

percentile cutoff values differ. Choice of birth weight standard significantly influences test characteristics. Negative prediction of large birth

weight is more accurate than positive prediction. At third trimester sonography with maternal diabetes, the abdominal circumference

percentile is potentially useful and should be routinely reported. D 2000 Elsevier Science Inc. All rights reserved.

Keywords: Fetus; Ultrasound; Diabetes mellitus; Abdominal circumference; Large for gestational age

1. Introduction

Almost 20 years ago, Ogata et al. [1] associated an

abnormally large fetal abdominal circumference (AC) mea-

sured between 28 and 32 weeks with `̀ accelerated somatic

growth'' in diabetic pregnancy. Fetuses with AC measure-

ments two standard deviations greater than the mean for

gestational age became newborns with increased birth

weight and increased skin-fold thickness. They also had

increased amnionic fluid insulin levels. Seven years later,

Bochner et al. [2] found that fetuses of diet-controlled

gestational diabetic women with AC greater than the 90th

percentile at 30±33 weeks had larger birth weight babies

with increased shoulder dystocia, birth trauma, and Cesar-

ean delivery for arrested labor. Subsequently, Buchanan et

al. [3] performed a controlled clinical trial in which women

with mild gestational diabetes and fetal AC greater than the

75th percentile at 29±33 weeks were randomly assigned to

insulin + diet vs. diet alone. Women in the insulin group

delivered babies with decreased skin-fold thickness who

were less than one-third as likely to be large for gestational

age (LGA) at birth. This demonstrated the potential utility of

fetal sonographic information to select diabetic women for

intensification of management.

These findings raised several issues for us. Some built-in

sonographic reporting systems do not routinely indicate a

numerical AC percentile, but the estimated fetal weight

(EFW) percentile is universally reported. Does the EFW

in the early third trimester also predict LGA birth weight?

There are various reference standards for fetal biometry and

for birth weight as functions of gestational age. Are predic-

tion characteristics similar with different reference stan-

dards? Finally, does timing of the ultrasound examination

within the early third trimester interval affect the relation-

ship between AC and subsequent birth weight? The answers

to these questions are not available in the literature to date.

The objective of this study is to examine the test

characteristics of early third trimester AC and EFW

as predictors of LGA birth weight using different

reference standards.

2. Materials and methods

From mid-1993 to mid-1995, patients at a university-

based specialty clinic for diabetic pregnancy routinely

received fetal sonography between 28 0/7 weeks and 31

* Corresponding author. Department of Obstetrics and Gynecology,

Division of Maternal± Fetal Medicine, St. Louis University, 6420 Clayton

Road, St. Louis, MO 63117, USA. Tel.: +1-314-768-8873; fax: +1-314-

768-8776.

0899-7071/00/$ ± see front matter D 2000 Elsevier Science Inc. All rights reserved.

PII: S0 8 9 9 - 7 0 7 1 ( 0 0 ) 0 0 1 53 - 4

Journal of Clinical Imaging 24 (2000) 1± 7

Page 2: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

6/7 weeks. The staff caring for these patients was committed

to tight glucose control. All patients were instructed to

check capillary glucose values at least four times per day.

Target glucose levels were fasting values less than 90 mg/dl

and 1-h postprandial values less than 140 mg/dl. The staff

attempted at least weekly contact with all patients regarding

glucose values and insulin adjustments. Gestational diabetes

was diagnosed according to the National Diabetes Data

Group criterion, except women with only one abnormal

value on 3-h glucose tolerance testing were also included

since they have also been shown to experience a high

frequency of excessive fetal growth [4,5].

At each scan, registered sonographers obtained fetal

biometry and a physician reviewed the images. Demo-

graphic and pregnancy outcome data were entered into a

computerized database designed specifically for the diabetes

program. Those with multiple gestations were excluded

from the study. AC estimates were obtained by summing

the anterior±posterior and the transverse axial diameter at

the level of the portal-umbilical venous complex and multi-

plying the sum by 1.57. AC percentiles were derived using a

reference standard published by Hadlock et al. [6] and also

using a reference standard reported by Bochner et al. [2].

The latter standard is based on a Los Angeles population

and was used in the studies of Bochner et al. [2] and

Buchanan et al. [3]. EFW percentiles were derived from a

standard published by Hadlock et al. [7], which utilizes an

EFW formula calculated from the head circumference, AC,

and femur length.

We determined the birth weight percentiles from two

different newborn standards: Brenner et al.'s [8] widely

utilized standard based on births in the Cleveland area and

that of Williams et al. [9] based on births in the state of

California. In studies of Bochner et al. [2] and Buchanan et

al. [3], the newborn weight standard of Williams et al. was

used to define LGA birth weight.

To evaluate the appropriateness for our population of the

AC percentile cutoff used by Buchanan et al. [3], we

constructed a receiver operating characteristic (ROC) curve

for AC percentile as a predictor of LGA birth weight. We

also constructed a ROC curve for EFW percentile predicting

LGA birth weight and selected a favorable cutoff value. We

then calculated sensitivity, specificity, positive predictive

value (PPV), and negative predictive value (NPV) for high

AC and EFW as predictors of LGA birth weight.

The stability of the relationship between the early third

trimester fetal AC and subsequent birth weight was assessed

by calculating the difference between birth weight and AC,

with both values adjusted for gestational age and expressed

as multiples of the median (momBWTÿmomAC). This

difference was then grouped according to the week (between

28 and 31 weeks, inclusive) that the sonogram had been

performed. The median values of (momBWTÿmomAC)

were compared among the four 7-day intervals. Multiples

of the median were used for gestational age adjustment

because both Brenner's and Williams' newborn weight

standards are tabular and non-parametric. After assuring a

normal distribution of momBWT, the linear regression of

momBWT on momAC was conducted and the residuals were

analyzed. The Cook±Wesiberg test [10] of heteroscedasti-

city (non-constant variance) and linear regression of the

residuals on gestational age at sonography were performed.

Fisher's exact test was used for comparison of propor-

tions and the Kruskall±Wallis test for comparison of med-

ians. The test of Cuzick was used for non-parametric test of

trend [11]. A p-value of <0.05 was considered as statisti-

cally significant. Stata (Stata, College Station, TX) was used

for statistical analysis.

3. Results

Ninety-two patients had ultrasound examinations be-

tween 28 and 32 weeks. Five patients with twins and three

patients without delivery outcome data were excluded leav-

ing 84 patients for analysis. The characteristics of these

patients are shown in Table 1. Six of the patients with

gestational diabetes had one abnormal value on 3-h glucose

tolerance testing; two of these required insulin therapy.

There are significantly more infants identified LGA by

the Brenner standard than by the Williams standard ( p =

0.028). All infants with LGA birth weights by the California

standard were also LGA by the Brenner standard. Twelve

infants that were LGA by Brenner's standard were not LGA

by the California standard. Among these 12, only one had a

birth weight above 4000 g (she weighed 4140 g). The

remaining 10 infants in the study with weights at or above

Table 1

Demographic characteristics of the study group

Age 29.31 � 6.83

Gravidity 3.00 � 1.79

Parity 1.06 � 1.23

Race (n)

African± American 40

White 40

Other 4

Diabetes (n)

Preexisting 42

Gestational, diet 18

Gestational, insulin 24

GA at ultrasound 30.26 � 1.09

GA at delivery 37.21 � 2.11

Birth weight 3252 � 712

LGA (n; %) (Brenner)

Preexisting 19 (45.2%)

Gestational 13 (31.0%)

All 32 (38.1%)

LGA (n; %) (Williams)

Preexisting 11 (26.2%)

Gestational 7 (16.7%)

All 18 (21.4%)

Values are expressed as mean � standard deviation unless otherwise

stated. Large for gestational age (LGA) birth weight is determined by two

reference standards, those of Brenner and Williams (see text).

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±72

Page 3: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

Fig. 1. An ROC curve for percentile abdominal circumference (Hadlock) to predict LGA birth weight (Williams). The asterisk indicates the 75th percentile

cutoff for abdominal circumference.

Fig. 2. An ROC curve for percentile abdominal circumference (Hadlock) to predict LGA birth weight (Brenner). The asterisk indicates the 75th percentile

cutoff for abdominal circumference.

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±7 3

Page 4: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

4000 g were LGA by both standards. In short, the California

standard sets higher weight thresholds for the 90th percen-

tile and identifies the largest of the large newborns as LGA.

The ROC curve for AC percentile (Hadlock) predicting

LGA birth weight (Williams) is in Fig. 1. The point

representing an AC cutoff at the 75th percentile is near

the point of maximal test efficiency, but favors high sensi-

tivity over specificity, as is appropriate for a test used to

select out a low risk group. With the Brenner standard for

LGA birth weight, area under the ROC curve is consider-

ably reduced (0.7124 vs. 0.8476; see Fig. 2). The ROC

curve for EFW percentile (Hadlock), in Fig. 3, demonstrates

a sharp decrement in sensitivity between the percentile

cutoff values of 70 and 75, and indicates the 70th percentile

as a more favorable cutoff for EFW.

The proportions of patients with fetuses above the 75th

percentile for AC are 0.429 for AC (Hadlock) and 0.381 for

AC (Bochner); the proportion above the 70th percentile for

EFW is 0.333. Test characteristics for these parameters are

in Table 2. For all AC and EFW reference standards, the

NPV (for predicting non-LGA birth weight) is significantly

higher with Williams' birth weight standard than with

Brenner's. Sensitivity is also higher when Williams' stan-

dard is used as opposed to Brenner's. The NPV is higher

than the PPV for all sonographic parameters, and this

difference is statistically significant with the Williams birth

weight standard. Early third trimester sonography is, thus,

best suited for ruling out subsequent delivery of the largest

of the large birth weight infants.

The relationship between momBWT and momAC, as a

function of the gestational age at ultrasound scan, is dis-

played in Fig. 4. Comparison of median values for

Fig. 3. An ROC curve for percentile estimated fetal weight (Hadlock) to predict LGA birth weight (Williams). The asterisk indicates the 70th percentile cutoff

for estimated fetal weight.

Table 2

Test characteristics for abdominal circumference (AC) and estimated fetal

weight (EFW) as predictors of LGA birth weight

LGA

(Brenner)

LGA

(Williams)

Sensitivity AC (Hadlock) 0.7001 0.9441

AC (Bochner) 0.633 0.889

EFW (Hadlock) 0.5312 0.8892

Specificity AC (Hadlock) 0.722 0.712

AC (Bochner) 0.759 0.758

EFW (Hadlock) 0.788 0.818

PPV AC (Hadlock) 0.600 0.4723

AC (Bochner) 0.594 0.5004

EFW (Hadlock) 0.607 0.5715

NPV AC (Hadlock) 0.8136 0.9793,6

AC (Bochner) 0.7887 0.9624,7

EFW (Hadlock) 0.7328 0.9645,8

Statistically significant comparisons are as follows: 1, p = 0.036; 2, p =

0.013; 3, 4, 5, p < 0.001; 4, 5, 6, p < 0.002. All other comparisons yield

statistically insignificant differences.

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±74

Page 5: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

(momBWTÿmomAC) among the four 7-day gestational

age intervals revealed no significant differences (c2 =

1.27; p = 0.736). In addition, a non-parametric test of trend

was negative (z = ÿ0.67; p = 0.51).

Examination of the scatter plot of momBWT on momAC

(not shown) reveals a clear linear relationship. Using the

method of Hadi and Simonoff [12], two outliers are identi-

fied with momBWT much less than expected for momAC.

Both of these individuals had hypertensive complications in

the third trimester that could explain a deceleration in fetal

growth. Omitting these two outliers, the linear relationship

between momBWT, and momAC is:

momBWT � 1:910�momAC� ÿ 0:890 �R2 � 0:472�:The variance of this relationship, reflecting random error,

changes as a function of gestational age at sonography

(Cook±Weisberg p = 0.002) with larger residuals at earlier

gestational ages (coefficient ÿ0.0263; p = 0.002). As

otherwise stated, there is no systematic trend in the

relationship between AC and subsequent birth weight,

but there is higher correlation when sonography is per-

formed at gestational ages later in the 28- to 32-week

interval. Wider scatter at earlier gestational ages is ob-

servable in Fig. 4.

4. Conclusions

Theoretical support for the use of fetal AC, as a

predictor of LGA growth in diabetic pregnancy, is well

summarized by Kehl et al. and others [13,14]. At birth,

LGA infants of diabetic mothers have 17% increased lean

body mass compared with the non-LGA infants, but the

increase is fat mass is a remarkable 99%. Third trimester

sonographic growth rates for AC, abdominal fat thick-

ness, thigh fat thickness, and liver length are significantly

greater for LGA compared with non-LGA fetuses; head

and femur growth rates are no different [14]. Others have

studied the sonographic evaluation of excessive fetal

growth in diabetic pregnancy, but no previous study has

focused on comparison of the AC and EFW early in the

third trimester when timely interventions may modulate

subsequent growth. Also, no previous study of this topic

has considered the effect of using other commonly

employed reference standards. We found that early third

trimester sonography yielded significantly better negative

than positive predictions for subsequent LGA birth

weight. Our results are comparable to those of Bochner

et al. [2] who found a sensitivity of 0.878, specificity of

0.825, PPV of 0.563, and NPV of 0.964 when a 90th

percentile cutoff AC was used between 30 and 33 weeks

in a group with mild gestational diabetes. Our values

were 0.889, 0.758, 0.500, and 0.962, respectively. Boch-

ner et al. evaluated only women with diet-controlled

gestational diabetes at a later gestational age interval,

and used a higher AC percentile cutoff value.

An earlier study by Tamura et al. [15] assessed the test

characteristics for fetal AC and estimated weight as

predictors of LGA birth weight in a diabetic population.

They found similar performance of the AC compared with

the EFW, as we did in our study. Tamura et al. [15] used

the Brenner birth weight standard and scans were per-

formed substantially closer to delivery at a mean gesta-

tional age of about 36 weeks. They used a 90th percentile

cutoff values for both AC and EFW. Sensitivity, specifi-

city, PPV, and NPV were 0.78, 0.81, 0.78, and 0.81,

respectively, for AC, and 0.78, 0.78, 0.74, and 0.81,

respectively, for EFW. The higher PPV in their study

Fig. 4. The difference between birth weight and sonographic AC, each expressed as multiples of the median (MOM), and plotted against gestational age at the

time of sonography in weeks. Medians are indicated by plus symbols.

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±7 5

Page 6: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

may be related to a later gestational age at sonography, a

higher percentile cutoff value, and a higher proportion of

LGA newborns (46% in their population compared with

38% in ours).

Another study relevant to ours is that by Johnstone et al.

[16] who performed sonography on 137 diabetic women

between 27 and 29 weeks gestation. Twenty-nine percent of

these women delivered macrosomic infants by the authors'

definition (greater than the 95th birth weight percentile by a

British standard). The test characteristics for AC were

disappointing with sensitivity, specificity, PPV, and NPV

of 0.28, 0.91, 0.55, 0.77, respectively. The percentile cutoff

value for AC was not stated by the authors and may have

been unusually high since only 15% of these diabetic

women had a high fetal AC.

We focused on the ability of a single early third trimester

sonogram to predict LGA growth because there is the

potential to modulate diabetes management and fetal sur-

veillance based on this information alone. Two other studies

are related to ours, but not directly comparable, since they

incorporated information from serial sonography and calcu-

lated growth rate to predict LGA birth weight. Rossavik et

al. [17] mathematically modeled fetal growth using sono-

graphic data throughout pregnancy to project birth weight.

In a preliminary study with 55 diabetic women, they found a

sensitivity of 1.00 and specificity of 0.98 for the detection of

LGA birth weight (greater than 90th percentile using the

Brenner standard). Landon et al. [18] employed a simpler

method of calculating the change in AC between serial scan

beyond 32 weeks. They found that a cutoff value of 1.2 cm/

week yielded a sensitivity of 0.84 and specificity of 0.85 for

detection of LGA birth weight (greater than 90th percentile

by Brenner's standard).

Both standards for the AC percentile (Hadlock's and

Bochner's) performed similarly in our population. The

pattern of test characteristics for early third trimester AC

would favor its use in situations where an intervention is

confined to the high-risk group and avoidance of false

negative predictions is particularly important. An AC cutoff

value at or near the 75th percentile seems optimal when high

sensitivity is desired. Test characteristics for the EFW were

quite similar to those of the AC in our population after

appropriate choice of a percentile cutoff. It cannot be

assumed that the optimal percentile cutoff is the same for

EFW as for AC. We found the 70th percentile to be a better

cutoff value for the EFW.

The reference standard for birth weight had a pronounced

effect on test characteristics of the AC and the EFW. The

most appropriate standard to use depends, in part, on the

racial and ethnic mix of the population cared for. It may also

depend on the goals of the clinician. If the object, for

instance, is to avoid the delivery of the baby greater than

4250 g, then a higher threshold standard, such as Williams'

California birth weight tables, may be the most useful.

Clearly, one could achieve a similar effect by using a

higher percentile cutoff with the Brenner standard. Since

all birth weight standards suffer some weaknesses in

construction, and since they differ significantly, our

results highlight some arbitrariness in using LGA birth

weight as a clinical outcome variable. The reader must be

sensitive to the weight standards employed, as well as

population differences, when comparing results from

different centers.

In this study, the association between AC and subsequent

birth weight was stable across the 28- to 32-week interval,

except that increased random error (i.e., poorer correlation)

was evident with earlier scans. Essentially, there was more

noise in the relationship between AC and subsequent birth

weight with scans close to 28 weeks than with scans close to

32 weeks. Most likely, the timing of ultrasound examination

within this interval is not critical for predicting LGA birth

weight. However, improved correlation at later gestational

ages should be associated with better predictions. On the

other hand, information that are available later in gestation

leaves less time and opportunity for the therapeutic effect

from an intervention.

This study has limitations. The study group is hetero-

geneous with both preexisting and gestational diabetic

women, and there are not sufficient patients to allow for

subgroup analysis. Both preexisting and gestational diabetic

women in this study had high frequencies of LGA infants

and, presumably, the basis for excessive fetal growth was

similar. Nonetheless, a larger study examining test charac-

teristics separately in preexisting and gestational diabetic

women would be of interest. Another limitation is the

potential effect of intensive maternal therapy (diet, insulin,

delivery timing) on the natural history of fetal growth. It is

possible that relationship between early third trimester

biometry and subsequent large birth weight would be

different in less intensively managed patients. If any-

thing, treatment would be expected to blunt the PPV and

the specificity.

The AC measurement is well standardized, familiar, and

rapidly obtainable. A large AC is not specific for diabetic

fetopathy since many other genetic and environmental

factors affect fetal growth. A reliable and reproducible

parameter more specific for abnormal growth in the fetus

of a diabetic mother would be desirable. Newer techniques,

such as three-dimensional sonography, may be explored for

this purpose. At present, the AC is a sensitive, and poten-

tially useful, predictor of LGA (or non-LGA) birth weight in

diabetic pregnancies. This information may affect decisions

regarding choice of therapy early in the third trimester or

subsequent timing of delivery. The AC percentile should

be reported, numerically and/or graphically, when the

diabetic woman has third trimester fetal sonography. In

the non-diabetic woman, an unexpectedly large fetal AC

could prompt diabetes screening, if it has not been done

recently. Results from this study may help the clinician

define the risk of LGA birth weight based on early third

trimester sonography, as well as assist investigators in the

design of future studies.

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±76

Page 7: Abdominal circumference vs. estimated weight to predict large for gestational age birth weight in diabetic pregnancy

References

[1] Ogata E, Sabbagha R, Metzger B, Phelps R, Depp R, Freinkel N.

Serial ultrasonography to assess evolving fetal macrosomia. JAMA

1980;243:2405± 8.

[2] Bochner C, Medearis A, Williams J, Castro L, Hobel C, Wade M.

Early third-trimester ultrasound screening in gestational diabetes to

determine the risk of macrosomia and labor dystocia at term. Am

J Obstet Gynecol 1987;157:703± 8.

[3] Buchanan T, Kjos S, Montoro M, et. al. Use of fetal ultrasound to

select metabolic therapy for pregnancies complicated by mild gesta-

tional diabetes. Diabetes Care 1994;17:275±83.

[4] National Diabetes Data Group. Classification and diagnosis of dia-

betes mellitus and other categories of glucose intolerance. Diabetes

1979;28:497±501.

[5] Langer O, Brustman L, Anyaegbunam A, Mazze R. The significance

of one abnormal glucose tolerance test value on adverse outcome in

pregnancy. Am J Obstet Gynecol 1987;157:758± 63.

[6] Hadlock F, Deter R, Harrist R, Park S. Estimating fetal age: computer-

assisted analysis of multiple fetal growth parameters. Radiology

1984;152:497±501.

[7] Hadlock F, Harrist R, Sharman R, Park S. Estimation of fetal weight

with the use of head, body, and femur measurementsÐa prospective

study. Am J Obstet Gynecol 1985;151:333± 7.

[8] Brenner W, Edelman D, Hendricks C. A standard of fetal growth for the

United States of America. Am J Obstet Gynecol 1976;126:555±64.

[9] Williams R, Creasy R, Cunningham G, Hawes W, Norris F, Tahiro M.

Fetal growth and perinatal viability in California. Obstet Gynecol

1982;59:624±32.

[10] Cook R, Weisberg S. Diagnostics for heteroscedasticity in regression.

Biometrika 1983;70:1±10.

[11] Cuzick J. A Wilcoxon-type test for trend. Stat Med 1985;4:87± 90.

[12] Hadi A, Simonoff J. Procedures for the identification of multiple out-

liers in linear models. J Am Stat Assoc 1993;88:1264± 72.

[13] Catalano P, Tyzbir E, Allen S, McBean J, McAuliffe T. Evaluation of

fetal growth by estimation of neonatal body composition. Obstet Gy-

necol 1992;79:46 ± 50.

[14] Kehl R, Krew M, Thomas A, Catalano P. Fetal growth and body

composition in infants of women with diabetes mellitus during preg-

nancy. J Matern± Fetal Med 1996;5:273± 80.

[15] Tamura R, Sabbagha R, Depp R, Dooley S, Socol M. Diabetic macro-

somia: accuracy of third trimester ultrasound. Obstet Gynecol

1986;67:828±32.

[16] Johnstone F, Prescott R, Steel J, Mao J, Chambers S, Muir N. Clin-

ical and ultrasound prediction of macrosomia in diabetic pregnancy.

Br J Obstet Gynaecol 1996;103:747± 54.

[17] Rossavik I, Torjusen G, Deter R, Reiter A. Efficacy of mathema-

tical methods for ultrasound examinations in diabetic pregnancy.

Am J Obstet Gynecol 1986;155:638± 44.

[18] Landon M, Mintz M, Gabbe S. Sonographic evaluation of fetal ab-

dominal growth: predictor of the large-for-gestational-age infant in

pregnancies complicated by diabetes mellitus. Am J Obstet Gynecol

1989;160:115± 21.

W.L. Holcomb Jr. et al. / Journal of Clinical Imaging 24 (2000) 1±7 7