Prediction of Birth Weight Using the RossavikGrowth Model: A Study in a Dutch Population

Yenny S. Kurniawan, MD,1,4 Russell L. Deter, MD,2

Gerard H.A. Visser, MD, PhD,3 Johan L. Torringa, MD, PhD4

1 Department of Obstetrics and Gynaecology, University Hospital, Groningen,The Netherlands

2 Baylor College of Medicine, Houston, Texas3 University Hospital, Utrecht, The Netherlands4 Department of Obstetrics and Gynecology, Martini Hospital, van Swietenlaan 4,

9728 NZ Groningen, The Netherlands

Received 6 February 1996; accepted 29 July 1996

Abstract: Objectives: To evaluate the Rossavik growth model for predicting birthweight in a Dutch population and to evaluate growth cessation near term.

Study Design: Birth weight was predicted at various ages between 38 and 42 weeks,menstrual age (MA), and at birth age in 50 normal infants using two sets of ultra-sound measurements obtained before 28 weeks, MA. Predicted birth weights werecompared to actual weights. The mean percentage difference was used as a measureof systematic error and the standard deviation as a measure of random error. Linearregression analysis was used to evaluate the relationship between percentage differ-ences and birth age. To evaluate the individual growth potential, the Growth PotentialRealization Index for weight (GPRIWT) was determined for each fetus.

Results: The predictions at 39 and 39.15 weeks, MA, were accurate without system-atic error and with a random error of 69.3%. Prediction at 38 weeks showed a statisti-cal underestimation (mean 6 SD 5 25.8% 6 8.8), and statistical overestimationswere found for predictions after 39.15 weeks and at birth age. A relationship betweenpercentage differences and birth age was not found for predictions between 39.15 and40 weeks, MA. These findings indicate that growth cessation occurred at 39.15 weeks,MA. Using birth weights predicted at 39.15 weeks, MA, GPRIWT were calculated. Themean GPRIWT value was not significantly different from 100% (p . 0.05), and individ-ual GPRIWT values ranged from 84% to 114%.

Conclusions: The Rossavik growth model can be used to predict birth weight in aDutch population. However, growth cessation near term appears to occur later thanpreviously reported in other populations. 1997 John Wiley & Sons, Inc. J Clin Ultra-sound 25:235–242, 1997Keywords: fetal growth; fetal weight; Rossavik growth model

In most studies of fetal growth, birth weight has Because there was no better method for evalu-ating intrauterine growth initially, birth weightbeen used in the evaluation of growth outcome at

birth.1–4 In making such evaluations, birth data at various menstrual ages were collectedand used to construct age-specific size curvesweight must be related to menstrual age, to par-

ity, and also to sex. (cross-sectional studies). Individual birth weightswere then compared to these standards. New-borns were classified as small-for-gestational age

Correspondence to: Y.S. Kurniawan at Martini Hospital, (SGA), appropriate-for-gestational age (AGA)Groningen, The Netherlands and large-for-gestational age (LGA). However,

there is no international agreement on the defi- 1997 John Wiley & Sons, Inc. CCC 0091-2751/97/050235-08

VOL. 25, NO. 5, JUNE 1997 235

KURNIAWAN ET AL

MATERIALS AND METHODSnition of SGA, AGA, and LGA.5 Some investiga-tors prefer to use the 10th percentile as the cut-

Patient Sampleoff point for SGA; others use the 5th percentile,the 2.3rd percentile, or 2SD below the mean This study was carried out in 50 of 81 nullipa-for menstrual age.5 The limitations of using rous women from the district of Groningen (Thethis cross-sectional approach are that growth re- Netherlands). These pregnancies were consid-tarded and macrosomic neonates cannot be sepa- ered ‘‘normal’’ based on five criteria: (1) tworated from genetically small and large neonates sets of ultrasound measurements, with an in-who are normal, and some AGA neonates are ac- terval of 4 to 8 weeks between scans and head/tually growth retarded or macrosomic.6–9

abdominal cube values within 63SD of meanAn alternative approach to the evaluation of values determined previously17; (2) uncompli-

growth outcome at birth is Individualized cated pregnancy; (3) delivery after 37.0 weeks;Growth Assessment, a procedure in which each (4) birth weight .2.3rd percentile; (5) normalfetus serves as its own control. As shown by De- pediatric assessment of the newborn. This pa-ter and Rossavik,10 fetal weight can be estimated tient sample has been described in detail pre-from head and abdominal cube measurements viously.15,16 All pregnancies were singletons andusing a function first described by Rossavik.8 Us- there were no congenital abnormalities, congen-ing predicted head and abdominal cube values ital infections, or maternal drug/alcohol abuse.derived from Rossavik growth models specified The MA was based on the first day of the lastfrom measurements obtained before 26 weeks, menstrual period confirmed by an ultrasoundmenstrual age (MA), birth weights in singletons examination [crown–rump length as describedcan be predicted with a mean accuracy of 0.2% by Robinson18 and/or biparietal diameter (BPD)(range: 27.7% to 18.9%).7 These findings are as described by Campbell19] at the first prenatalsupported by the results presented by Simon et visit between 6 and 15 weeks, MA, in 46 cases.al.11 Individualized Growth Assessment provides In 4 cases, fetal age was determined by ultra-correction for differences in growth potential and sound because of previous irregular menstrualthus permits comparisons of the actual birth cycles. Smoking was reported in 35% of theweight to the expected birth weight if the fetus woman, all fewer than 6 cigarettes a day.has grown normally. This comparison can be ex-pressed as a ratio of the actual weight to the pre-

Ultrasound Examinationdicted weight multiplied by 100, which has beencalled the Growth Potential Realization Index for Ultrasound examinations were carried out at ap-weight (GPRIWT).12

proximately 17 weeks (mean: 17.2 weeks, range:In singleton neonates with normal growth out- 14.7–19.7 weeks) and 25 weeks (mean: 25.3

comes, previous studies have indicated that the weeks, range: 22.1–28.0 weeks) by one investiga-GPRIWT has a mean value of 100% and a range tor (YSK) using dynamic image ultrasound scan-of 92% to 108% in an American population,5,7,12

ners (Aloka 250, Acuson 600, sound velocity 1540while in a Japanese population, a mean value of m/s, 3.5 MHz transducers). BPD and HC were98% and a range of 89% to 114% was found.13

determined in a section containing the thalamus:Studies concerning weight prediction at term for the BPD was measured from outer to inner mar-American populations have demonstrated that gins. Assuming the head profile shape to be anthe weight does not increase after 38 weeks, MA, ellipse, the HC was calculated using the largestindicating a cessation of growth.12,14 In our previ- (HLA) and the shortest (HSA) axes obtainedous investigations of a Dutch population,15,16 we from the outer contour of the skull.20

found that growth parameters, such as the head The abdominal circumference (AC) was calcu-circumference (HC) and the crown–heel length lated using the largest (ALA) and the shortest(CHL), appear to stop growing at about 39.5 (ASA) axes obtained from the outer contour ofweeks, MA. the abdomen in the reference plane perpendicu-

Our purpose was to evaluate this approach to lar to the fetal spine which shows the umbilicalweight prediction in a Dutch population to deter- vein.20 Because the diameter measurements weremine if this method can also be used for Dutch not recorded at the time of data collection, thefetuses/neonates. A re-evaluation of the presence HLA and HSA were subsequently estimatedand timing of growth cessation near term was from the BPD and HC and the ALA and ASA

from the AC using functions described by Deteralso part of our study.

236 JOURNAL OF CLINICAL ULTRASOUND

PREDICTION OF BIRTH WEIGHT

TABLE 1 and B values as described by Deter et al10 (Ta-Functions Used to Obtain Estimates of Fetal Parameters ble 2).

‘‘HSA’’ 5 0.18052 1 1.03523 (BPD) The predicted birth weights were compared to‘‘HLA’’ 5 0.979 3 Ï[(HC2/4.9298) 2 HSA2] the actual weight measurements at birth. When‘‘ASA’’ 5 0.33777 1 0.84561 (AC/3.14) a fetus was born before a specified age, its pre-‘‘ALA’’ 5 20.09336 1 1.05462 (AC/3.14)

dicted weight at its birth age was used. The per-A 5 (‘‘HSA’’ 3 ‘‘HLA’’)1.5

B 5 (‘‘ASA’’ 3 ‘‘ALA’’)1.5 centage difference [(predicted 2 actual)/actual 3‘‘HSA’’ 5 estimated head short axis; ‘‘HLA’’ 5 estimated head 100] was determined for each neonate. The mean

long axis. and standard deviation of the percentage differ-BPD 5 biparietal diameter; HC 5 head circumference.

ences were calculated: the mean value was taken‘‘ASA’’ 5 estimated abdominal short axis; ‘‘ALA’’ 5 estimated ab-

dominal long axis. as a measure of the systematic prediction errorAC 5 abdominal circumference. and the standard deviation as a measure of theA 5 head cube.

random prediction error. Linear regression anal-B 5 abdominal cube.

ysis was used to evaluate the relationship be-tween percentage differences and birth age. Themean difference was compared to zero using the

et al.6 The head cube and the abdominal cube Student t-test to evaluate the systematic error;were determined using the equations given in the variability accounted for by the regressionTable 1. was evaluated by the R2 value.

To evaluate growth taking into account indi-vidual growth potentials, growth potential real-

Postnatal Measurements ization index value for weight (GPRIWT 5 actualbirth weight/predicted birth weight 3 100) wasMeasurements obtained for the neonate weredetermined for each fetus using the predictedtaken immediately after delivery: weight inbirth weight at the optimal growth cessationgrams, CHL and HC in centimeters, Apgarage.12 The mean and SD for the GPRIWT valuesscores, and umbilical artery and vein pH values.were calculated and the former compared withA neonate was categorized as normal if the pedi-100% using the Student t-test. A p value ,0.05atric assessment of the infant was normal (nor-was taken as an indication of statistical sig-mal appearance at birth, no congenital abnor-nificance.malities, Apgar scores above 7 after 1 and 5

minutes). The mean birth age was 40.3 weekswith a range from 38.3 to 42.0 weeks.

RESULTS

The mean (6SD, 100% range) for birth weightsData Analysis of the 50 infants of this normal subgroup was

3377 g (6339 g, 2740 g to 4030 g). Table 3 showsIndividual Rossavik growth models were deter-the values of the mean (6SD) and the 95% rangemined for each of the 50 fetuses.17,21 The generalof the percentage differences for neonatal weightequation of this model is

P 5 c(t)k1s(t)

TABLE 2where P is the anatomic parameter, k is a fixed Weight Prediction Procedure Using the Rossavik

coefficient related to the anatomy of the parame- Growth Model

ter studied, c is a variable coefficient related toSlope 5 (P2 2 P1)/(t2 2 t1)

the genetic regulators of growth, s represents an Start point 5 2[P2 2 (slope 3 t2)I/slope

Log(e) c 5 a0 1 a1 3 log(e)slopeunknown regulatory system that modifies geneti-s 5 b0 1 b1 3 coeff.ccally determined growth,22 and t is the duration

Parameter k a0 a1 b0 b1of the growth of the parameter.Values for start points and coefficients c and s Head cube (A) 3.856 20.5197 4.0876 20.011177 20.56226

Abdominal cube (B) 3.284 0.1344 3.6820 0.001768 20.22800for the head and abdominal cubes were deter-mined for each fetus from slope values calculated A 5 ca(ta)

ka1sa(ta)

B 5 cb(tb)kb1sb(tb)from data obtained in two scans before 28 weeks,

FW 5 0.46138 3 [1.337(ta)(20.04110.009173ta)] 3 AMA.17 Head and abdominal cubes (A, B) were

1predicted at various ages between 38 and 40 0.55029 3 [3.517(tb)

(20.07420.000743tb)] 3 B

weeks and at birth age. The predicted birth ta 5 menstrual age at start point of A.

tb 5 menstrual age at start point of B.weights were calculated from these predicted A

VOL. 25, NO. 5, JUNE 1997 237

KURNIAWAN ET AL

TABLE 3

Prediction of Birth Weights at Various Assumed Growth Cessation Ages

Between 38 and 40 Weeks and at Birth Age

AssumedRegressionGrowth

Percentage DifferencesCessation Slope ofAge Mean (6SD) Range t-test p Regression r2

(weeks) (%) (%) (%)

38 25.8 (68.8) 219.2 to 110.9 a a Neg 12.2

39 11.5 (69.3) 213.0 to 117.6 NS a Neg 8.7

39.15 12.3 (69.3) 212.3 to 118.7 NS NS None 7.2

39.25 12.7 (69.3) 213.5 to 119.1 a NS None 7.1

39.5 14.7 (69.5) 29.3 to 122.3 a NS None 3.1

40 17.4 (69.9) 27.4 to 126.9 a NS None 0.1

Birth age 111.8 (610.5) 26.8 to 128.5 a a Pos 15.7

aSignificant, p , 0.05; NS 5 not significant, p . 0.05.

Range includes 95% of the values.

N 5 50.

predictions at 38, 39, 39.15, 39.25, 39.5, 40 suggesting that the optimal growth cessation agefor weight was 39.15 weeks (Figure 1).weeks, MA, and at birth age. The prediction at

38 weeks, MA, showed a significant underestima- The mean value of the percentage differencesfor prediction at 39.15 weeks, MA, given in Tabletion. The prediction at 39 weeks was the most

accurate [no statistically significant systematic 3 showed 12.3% overestimation, but this valuewas not significantly different from zero as deter-error, mean (6SD) 5 1.5 (69.3)%, p . 0.05],

whereas the predictions after 39.25 weeks, MA, mined by the Student t-test. The random errorwas 69.3% and the 95% range was from 213%resulted in a significant overestimation; this sug-

gests growth cessation between these ages. Re- to 119%. When compared to the results obtainedin similar studies for American populations, wegression analysis showed no relationship be-

tween the percentage difference and birth age for found either similar (Simon) or greater (Deter)systematic and random errors (Table 4).the predictions between 39.15 and 40 weeks, MA,

FIGURE 1. Relationship between the difference between predicted and actual birth weight (% difference) and

birth age using a prediction age of 39.15 weeks (dotted line is 0% difference).

238 JOURNAL OF CLINICAL ULTRASOUND

PREDICTION OF BIRTH WEIGHT

TABLE 4

Comparison of Birth Weight Predictions in the Present Study with Those in Similar

Studies of Deter et al12,23 and Simon11

Comparison ofPercentage Differences

Mean to ZeroN Mean SD Rangea by t-testb

(%) (%) (%)

Kurniawanc 50 12.3 69.3 212.3 to 118.7 NS

Simon11 58 22.8 68.5 218.5 to 111.0 a

Deter12 20 11.0 Not reported 27.7 to 110.1 NS

Deter23 25 21.7 Not reported 213.6 to 19.6 NS

aRange includes 95% of the values.bNS 5 not significant (p . 0.05).cPredicted weights used in this study were those determined at 39.15 weeks, MA, while those in the

studies of Deter and Simon were those determined at 38 weeks, MA.dSignificant (p , 0.05).11Percent deviations were used rather than percent difference, so the sign is in opposite direction.

The sample of Simon et al11: Of the 70 infants, 58 were delivered between 27.1 wk and 37.9 wk. 80%

of the women were of low socioeconomic status, 80% were Caucasian, and 19 had risk factors or

complications that may affect fetal growth. All neonates had birth weights within 10th–90th percentile

(York BW curve).12Ultrasound measurements carried out at 2 to 3 week intervals, regression analysis based on data

obtained before 26.1 weeks were used to determine the values of start point, slopes, coeff. c and s.13The sample of Deter et al23: all singleton, primarily Caucasian, middle-class women, all neonates

had birth weights within the normal ranges (mean 62SD, weight table: Usher and McLean). Weights

were predicted using two ultrasound measurements obtained before 28 weeks, MA.

Based on the percentage differences using value was not significantly different from 100%(p 5 0.24), indicating that the fetuses had grownweight predictions at 39.15 weeks, 14 (28%) in-

fants had percentage differences greater than according to their individual growth curves. Ta-ble 6 shows the GPRI values for the three param-110%. Of this group, smoking of cigarettes (less

than 6 a day) was found in 8 (57%) women. Of eters studied in our population.the other 36 infants with percentage differencesless than 110%, smoking was found in 12 (33%)

DISCUSSIONwomen. This difference is not significant (p .0.05, Fisher’s exact test), suggesting that smok- Differences in growth patterns around term

might be caused by genetic factors and/or influ-ing of a small number of cigarettes was not theexplanation for the increased differences be- enced by environmental factors or by the social

environment as was recently demonstrated bytween the predicted and actual birth weights.The values of the GPRIWT based on birth Doornbos et al.24 It has been suggested that in

fetuses who were small and continued to growthweights predicted at 39.15 weeks, MA, are givenin Table 5. Our values for GPRIWT were almost at term, spontaneous labor started later because

they needed first to reach a critical size limit.12the same as those found by Deter et al,12 Hata etal,7 and Ariyuki et al13 using birth weights pre- On the other hand, cessation of growth might be

a biological process which allows the fetus to pre-dicted at 38 weeks, MA (Table 5). Our mean

TABLE 5

The Values of GPRIWT in Normal Fetuses

Comparison ofMean to 100%

N Mean SD Rangea by t-test

(%) (%) (%)

Hata7 99 100 Not reported 92–108 NS

Deter12 20 99.4 Not reported 91–108 NS

Ariyuki13 120 98 67.0 86–114 Not reported

Kurniawan 50 98.5 68.8 84–114 NS

aRange includes 95% of the values.

N 5 number of patients.

NS 5 not significant (p . 0.05).

VOL. 25, NO. 5, JUNE 1997 239

KURNIAWAN ET AL

TABLE 6 found, which supports a growth cessation age ofThe GPRI Values for Head Circumference (HC), 38 weeks in American populations. When using

Crown–Heel Length (CHL), and Weight (WT) Using38 weeks, MA, as growth cessation age, the Ros-Predicted Values at Optimal Growth Cessation Agessavik growth model failed to accurately predict

GPRI fetal weight at term in a Dutch population. De-Comparison of termination of the growth cessation age specific

Anatomic Mean to 100% for a population is, therefore, needed for the ap-Parameter Mean 6SD Range by t-test

plication of the Rossavik growth model in this(%) (%) (%) population.

HC16 100.3 61.9 97–103 NSIt is unclear why growth in the present DutchCHL15 99.2 63.2 93–104 NS

WT 98.5 68.8 84–114 NS population continued for 1 to 1.5 weeks longerthan in the American populations studied pre-NS 5 Not significant, p . 0.05.

Range includes 95% of the values. viously. In the Netherlands (and also in ourstudy population) all pregnant women stop work-ing outside of the home at 34 weeks without fi-nancial consequences. This may well have beendifferent in the American populations. However,pare for life after birth. This latter process could

result in pathology if spontaneous labor fails to Zuckerman et al25 did not find a relationship be-tween the maternal work history and neonataloccur too long after growth has stopped.

As confirmed in Table 3, use of prediction ages outcome in an American population in a differentsociodemographic group. Within the context ofabove 39.15 weeks and birth age resulted in an

increase in the systematic prediction error that longer fetal growth near term in our population,it is interesting that the average birth weight iswas significantly above zero. Such results indi-

cate that predicted values will be too high for fe- higher in the Netherlands than in the UnitedStates.26,27 It is possible that the soft tissue com-tuses delivering after that age.

The absence of a statistically significant rela- position of the Dutch babies differs from that ofthe babies studied previously. Hata et al9 havetionship between the percentage difference and

birth age (Figure 1) showed that cessation of fe- pointed out that the Rossavik weight estimationfunction tacitly assumes the ‘‘average’’ amount oftal weight growth occurs at about 39.15 weeks,

MA. Previously we found in the same data set the soft tissue present in the sample used to de-rive the weight estimation equation.that growth of both the crown–heel length (cal-

culated from predicted femur diaphysis length) The differences in weight prediction at the op-timal growth cessation age between our studyand the head circumference appeared to stop at

about 39.5 weeks, MA.15,16 Thus there is increas- and those previously reported for American pop-ulations were small (Table 4). The systematicing evidence for cessation of growth between 39

and 39.5 weeks in Dutch fetuses. The difference and random errors were either the same orslightly larger in our sample. Studies of weightin the cessation age among the three growth pa-

rameters is too small to conclude that cessation prediction using the Rossavik model showed thatonly Simon et al11 and Ariyuki et al13 found a sig-of weight growth occurs before that of the other

parameters and it is probably only of statistical nificant overestimation. This might be due to thecomposition of their study samples, becauseinterest.

Rossavik et al14 used a different method for de- women with complicated pregnancies were in-cluded. The small ranges found by Deter et al12termining the timing of growth cessation around

term. They based their evaluation on the absence might be due to the small size of the study sam-ple used but could also be explained by the differ-of changes in the mean percentage differences for

a certain age onward and found that from 38 ent method used: individual weights were pre-dicted using the start point, slope, and coefficientweeks, MA, onward, the mean percentage differ-

ences were not significantly different from each c and s values determined by regression analysisbased on data from more than two time pointsother; they concluded that growth cessation oc-

curs at about 38 weeks. Deter et al12 used the before 26.1 weeks, MA. Prediction of the weightusing this method might be more accurate thansame method as described in the present study

but only for the prediction at 38 weeks. However, the two-scan method because less variability wasfound in the prediction of abdominal and headthe mean prediction error was not significantly

different from zero, and no relationship between cubes.17 This may also explain the small range inGPRIWT values (Table 5).the percentage difference and birth age was

240 JOURNAL OF CLINICAL ULTRASOUND

PREDICTION OF BIRTH WEIGHT

mester growth and birth characteristics. ObstetOur mean GPRIWT value was the same as thatGynecol 78:379–384, 1991.found in American and Japanese populations

8. Rossavik IK: Patterns and principles of fetal(Table 5). However, the predicted values for thegrowth. Thesis, University of Oslo, Oslo, Norway,weight in the Japanese study was multiplied by1982, pp 167–168.0.963 to compensate for the systematic error

9. Hata T, Deter RL, Hill RM: Reduction of soft tis-(4.7%), which was not found in the present study sue deposition in normal triplets. J Clin Ultra-nor in the studies of American populations.12,27

sound 19:541–546, 1991.This systematic overestimation of the weight 10. Deter RL, Rossavik IK, Harrist RB: Developmentcould be due to the use of an inappropriate of individual growth curve standards for esti-growth cessation age (ie, 38 weeks, MA). Simon mated fetal weight: I. Weight estimation proce-

dure. J Clin Ultrasound 16:215–225, 1988.et al11 found a systematic underestimation of the11. Simon NV, Deter RL, Shearer DM, Levisky JS:weight in their study of weight prediction using

Prediction of normal fetal growth by the Rossavikthe Rossavik method; their growth cessation agegrowth modeling using two scans before 27 weeks,was 38 weeks, MA. Although all of these differ-menstrual age. J Clin Ultrasound 17:237–243,ences are small, they suggest that an examina-1989.tion of the time of growth cessation is indicated

12. Deter RL, Hill RB, Tennyson LM: Predicting thein specific populations.birth characteristics of normal fetuses 14 weeks

In conclusion, the results of this study and before delivery. J Clin Ultrasound 17:89–93, 1989.those reported earlier for CHL15 and HC16 indi- 13. Ariyuki Y, Hata T, Kitao M: Evaluation of perina-cate that with the appropriate growth cessation tal outcome using individual growth assessment:age, Individualized Growth Assessment can be comparison with conventional methods. Pediatricsused to predict the birth characteristics of 96:36–42, 1995.

14. Rossavik IK, Deter RL, Wasserstrum N: Mathe-normal Dutch neonates. The predicted birthmatical modeling of fetal growth: V. Fetal weightcharacteristics provide individualized stan-changes at term. J Clin Ultrasound 16:9–15, 1988.dards for evaluating the growth status of such

15. Kurniawan YS, Deter RL, Visser GHA, et al: Pre-neonates.diction of the neonatal crown–heel length from thefemur diaphysis length measurements. J ClinUltrasound 22:245–252, 1994.

REFERENCES 16. Kurniawan YS, Deter RL, Visser GHA: Predictedversus actual head circumference at birth: a study1. Mills MD, Nageotte MP, Elliot JP, et al: Reliabil-in a Dutch population using the Rossavik growthity of ultrasonographic formulary in the predictionmodel. Ultrasound Obstet Gynecol 5:123–128,of real weight and survival of very-low-birth-1995.weight infants. Am J Obstet Gynecol 163:1568–

17. Deter RL, Rossavik IK: A simplified method for1574, 1990.determining individual growth curve standards.2. Verloove-Vanhorick SP, Verwey RA: Project onObstet Gynecol 70:801–806, 1987.preterm and small for gestational age infants in

The Netherlands 1983. PhD thesis, University of 18. Robinson HP: Sonar measurement of fetal crown–rump length as means of assessing maturity inLeiden, 1987.

3. Simon NV, Levinsky JS, Shearer DM, et al: Influ- first trimester of pregnancy. BMJ 4:28–31, 1973.19. Campbell S: Ultrasonic fetal cephalometry duringence of fetal growth patterns on sonographic esti-

mation of fetal weight. J Clin Ultrasound 15:376– the second trimester of pregnancy. J Obstet Gynae-col Brit Cwlth 77:1057, 1976.383, 1987.

4. Sabagga RE, Minogue J, Tamura RK, Hungerford 20. Kurjack A, Breyer B: Estimation of fetal weight byultrasonic abdominometry. Am J Obstet GynecolSA: Estimation of birth weight by use of ultra-

sonographic formulas targeted to large-, appro- 125:962, 1976.21. Rossavik IK, Deter RL: Mathematical modeling ofpriate-, and small-for-gestational-age fetuses. Am

J Obstet Gynecol 160:854–862, 1989. fetal growth: I. Basic principles. J Clin Ultrasound12:529–533, 1984.5. Deter RL, Harrist RB, Hadlock FP, Carpenter RJ:

The use of ultrasound in the detection of intra- 22. Rossavik IK, Deter RL, Hadlock FP: Mathematicalmodeling of fetal growth: III. Evaluation of headuterine growth retardation: a review. J Clin Ultra-

sound 10:9–16, 1982. growth using the head profile area. J Clin Ultra-sound 15:23–30, 1987.6. Deter RL, Harrist RB, Hill RM: Neonatal growth

assessment score: a new approach to the detection 23. Deter RL, Rossavik IK, Carpenter RJ: Develop-ment of individual growth standards for estimatedof intrauterine growth retardation in the newborn.

Am J Obstet Gynecol 162:1030–1036, 1990. fetal weight: II. Weight prediction during the thirdtrimester and at birth. J Clin Ultrasound 17:83–7. Hata T, Deter RL, Hill RM: Individual growth

curve standards in triplets: prediction of third tri- 88, 1989.

VOL. 25, NO. 5, JUNE 1997 241

KURNIAWAN ET AL

24. Doornbos JPR, Nordbeck HJ, Van Enk AE, et al: 26. Voorhorst FJ, Puyenbroek JL, Robertson EA, etal: Verschillen de geboortegewichten van vroegerDifferential birthweights and the clinical rele-

vance of birthweight standards in a multiethnic en nu? Ned Tv Geneeskunde 134:998–1002,1990.society. Int J Gynaecol Obstet 34:319–324, 1991.

25. Zuckerman BS, Frank DA, Hingston R, et al: Im- 27. Brenner WE, Edelman DA, Hendricks CH: A stan-dard of foetal growth of the United States ofpact of maternal work outside home during preg-

nancy on neonatal outcome. Pediatrics 77:459, America. Am J Obstet Gynecol 126:555–564,1976.1986.

242 JOURNAL OF CLINICAL ULTRASOUND