Prediction of intrauterine growth restriction with customisedestimated fetal weight centiles

Philip Owena,*, Jo Ogaha, Lucas M. Bachmannb,c, Khalid S. Khanb

Objectives 1.To determine the value of third trimester customised estimated fetal weight centile ranking in theprediction of infants born with anthropometric features of intrauterine growth restriction (IUGR). 2. Tocompare the performance of customised centiles with fetal growth velocity in the prediction of IUGR.

Design Prospective, observational study.

Setting Department of Obstetric Ultrasound.

Population Two hundred and seventy-four women with low risk pregnancies who were participants in apreviously published longitudinal study of fetal biometry.

Methods Third trimester estimated fetal weight customised centiles were obtained after adjustment forgestational age at the time of ultrasound, birth order and gender, maternal weight, height and ethnic origin.Fetal growth velocity was calculated using the increment in the fetal abdominal area over a mean 28-dayinterval. Fetal abdominal area growth velocity was expressed as a standard deviation (Z score). Threeneonatal anthropometric measures were used to define IUGR: subscapular or triceps skinfold thickness<10th centile, ponderal index <25th centile, mid-arm circumference to occipito-frontal circumference <�1SD. The ability of estimated fetal weight customised centiles <5th centile and <10th centile to identifygrowth restriction was determined by calculating likelihood ratios.

Main outcome measure Likelihood ratios for the prediction of neonatal anthropometric features of IUGR.

Results Two hundred and fifty-eight infants had an estimated fetal weight customised centile calculated andone or more anthropometric measurements. The mean customised estimated fetal weight centile for caseswith and without a low skinfold thickness were 16.4 vs 41.7 ( P < 0.01); the mean values of customisedestimated fetal weight centile for cases with and without ponderal index <25th centile were 22.8 vs 42.7 ( P< 0.01); mean estimated fetal weight customised centiles for cases with and without mid-arm circumferenceto occipito-frontal circumference <�1 SD were 26.3 vs 40.5 ( P > 0.05). A customised estimated fetalweight centile of 5 or less had likelihood ratios (95% CI) of 4.9 (2.7–6.3), 6.8 (4.5–10.6) and 6.3 (3.7–14)for skinfold thickness <10th centile, ponderal index <25th centile and mid-arm circumference to occipito-frontal circumference <�1 SD, respectively. An estimated fetal weight customised centile of 10 or less hadlikelihood ratios of 4.5 (2.6–8.6), 4.1 (2.5–7.2) and 7.1 (3.5–24) for skinfold thickness <10th centile,ponderal index <25th centile and mid-arm circumference to occipito-frontal circumference <�1 SD,respectively. In the prediction of a ponderal index <25th centile, the fetal abdominal area velocitylikelihood ratio for a positive test is higher than the likelihood ratio for the 10th estimated fetal weightcustomised centile ( P ¼ 0.04) but is not significantly higher for the other outcomes. The fetal abdominalarea velocity likelihood ratio is not significantly higher than the likelihood ratios for the fifth estimated fetalweight customised centile for any of the three outcomes.

Conclusion Customised estimated fetal weight centiles in the late third trimester are moderately useful in theidentification of infants with IUGR but are less accurate than calculated growth velocity in the prediction ofan infant with a low ponderal index.

INTRODUCTION

The growth restricted infant is characterised by a reduc-

tion in subcutaneous fat and a reduced birthweight in

relation to its length; measurement of these parameters

more usefully identifies infants experiencing adverse peri-

natal outcome than does population-based birthweight

centiles1,2. Ultrasound fetal biometry is the cornerstone of

fetal growth assessment but single estimates of fetal size

are poor at identifying growth restricted infants3,4. Quan-

tifying serial fetal biometry is inevitably adversely influ-

enced by issues of measurement reproducibility and between

measurement time interval5,6, but despite this, growth veloc-

ity has been found to be useful in identifying the growth

restricted infant4,7.

Adjusting birthweight for maternal height, weight, par-

ity, ethnic group, gestational age at delivery and gender

BJOG: an International Journal of Obstetrics and GynaecologyApril 2003, Vol. 110, pp. 411–415

D RCOG 2003 BJOG: an International Journal of Obstetrics and Gynaecology

doi:10.1016/S1470-0328(03)02971-9 www.bjog-elsevier.com

aNorth Glasgow NHS University Trust, UKbBirmingham Women’s Hospital, UKcHorten Centre, Zurich University, Switzerland

* Correspondence: Dr P. Owen, Department of Obstetrics, Princess

Royal Maternity Hospital, Alexandra Parade, Glasgow, UK.

improves the ability of birthweight to identify infants with

growth restriction8 and also those experiencing adverse

perinatal outcome9. However, birthweight centiles cannot

be used to influence obstetric management because the

information is not available to the obstetrician. Single

ultrasound estimated fetal weights can be customised to

account for physiological variables, to identify the sub-

sequent delivery of a growth restricted infant. Direct

comparison of the value of such customised measurements

with that of fetal growth velocity measurements has not

previously been reported. This study explores and com-

pares the ability of the two methods of predicting intra-

uterine growth restriction (IUGR) in a single data set.

METHODS

Three hundred and thirteen women attending the ante-

natal clinic at Ninewells Hospital, Dundee were enrolled

into a study of longitudinal ultrasound fetal biometry; this

study has previously been described in detail10.

Women underwent ultrasound fetal biometry at two

weekly intervals throughout the third trimester; fetal mea-

surements included the biparietal diameter, fetal abdominal

area and fetal femur length using standard techniques10. All

measurements were made with an Aloka SSD 650 ultra-

sound machine by one of the authors (PO). Estimated fetal

weight was calculated using a previously validated formula

from the measurements of biparietal diameter, fetal abdom-

inal area and fetal femur length11. The last estimated fetal

weight prior to delivery was entered into an online cal-

culator available at http://www.wmpi.net together with

details of maternal height, weight, ethnic origin and birth

order, fetal gender and gestational age to obtain a custom-

ised estimated fetal weight centile12.

Fetal abdominal area velocity was calculated from the

increment between the last and third from last fetal abdom-

inal area measurement and converted to a standard devi-

ation (Z) score using previously published gestational age

specific mean and standard deviation values10. The cal-

culation of growth velocity and its ability to identify IUGR

in this population has previously been described; neonatal

nutritional status was estimated from measurements of

skinfold thickness, ponderal index and mid-arm circumfer-

ence to occipito-frontal circumference7.

The performance of customised estimated fetal weight

centiles in the prediction of neonatal growth restriction was

determined for cutoffs at both the 5th and 10th estimated

fetal weight customised centiles. We generated 2 � 2 tables

which led to a nominal scale of measurement with yes/no

units for IUGR for both the test under study (customised

estimated fetal weight) and the reference tests. The 2 � 2

tables were used to compute predictive validity using like-

lihood ratio, a clinically useful measure of test accuracy.

Likelihood ratios were similarly calculated for fetal abdom-

inal area velocity after a suitable Z score cutoff was chosen

from the receiver–operator characteristics curve.

This approach enables one to quantify the effect a

particular test result has on the probability of the outcome

using a simplified form of Bayes’ theorem:

Posterior odds ¼ ðprior oddsÞðlikelihood ratioÞ;where odds ¼ probability/(1 � probability) and probability

¼ odds (odds þ 1).

For a positive test result, the greater the likelihood ratio

then the more significant is the change in the probability of

growth restriction. Conversely, for a negative result, the

smaller the likelihood ratio, the more significant is the

probability of the absence of growth restriction13.

To compare the diagnostic accuracy of estimated fetal

weight customised centiles with that of fetal abdominal

area velocity, the paired positive likelihood ratios were

compared using the jacknife resampling technique14. The

quotient of the likelihood ratios was tested for statistical

significance against one using the one sample t test. Sta-

tistical analysis was performed using Excel (Office 2000,

Table 1. Test performance of customised estimated fetal weight centiles in the prediction of three measures of neonatal nutritional status.

Test Outcome Area under

receiver– operator

characteristics curve

Sensitivity

(%)

Specificity

(%)

Positive

prediction (%)

Negative

prediction (%)

Customised estimated

fetal weight <5th centile

Skinfold thickness <10th centile 0.81 20 96 43 89

Customised estimated

fetal weight <5th centile

Ponderal index <25th centile 0.74 19 97 54 87

Customised estimated

fetal weight <5th centile

Mid-arm to occipito-frontal

circumference ratio <�1 SD

0.7 25 96 33 94

Customised estimated

fetal weight <10th centile

Skinfold thickness <10th centile 0.8 47 88 38 92

Customised estimated

fetal weight <10th centile

Ponderal index <25th centile 0.74 42 90 41 90

Customised estimated

fetal weight <10th centile

Mid-arm to occipito-frontal

circumference ratio <�1 SD

0.7 56 88 27 96

412 P. OWEN ET AL.

D RCOG 2003 Br J Obstet Gynaecol 110, pp. 411–415

Microsoft, Redmond, Washington, USA) and Stata software

package (version 7.0, StataCorp. 1999, Stata Statistical

Software: Release 7.0 College Station, Texas, USA). We

compared the positive likelihood ratios of the 5th and 10th

estimated fetal weight customised centiles with fetal abdom-

inal area velocity for all three measures of neonatal nutri-

tional status.

RESULTS

Two hundred and seventy-four women continued in the

study. Two hundred and sixty women delivered at 37 weeks

or greater. Twenty-two (8%) and 11 (4%) infants had ad-

justed birthweights below the 10th and 3rd centiles, res-

pectively. Both estimated fetal weight customised centile

and one or more measurements for skinfold thickness,

ponderal index and/or mid-arm circumference to occipito-

frontal circumference ratio were available for 258 cases and

these cases were analysed further. Of these 258 cases with

at least one anthropometric measurement, 30 infants had

one or both skinfold thickness measurements below the

10th centile, 36 had a ponderal index below the 25th cen-

tile and 16 had mid-arm circumference to occipito-frontal

circumference <�1 SD. In some cases, anthropometric data

were not available because of early discharge or because

the reference data did not extend to the earlier gestational

ages.

The mean and range of estimated fetal weight custom-

ised centiles was 39.3 (0.6–99.8); the mean and range of

interval between gestational age at calculation of the

estimated fetal weight customised centiles and delivery

was 12 days (0–60). Some cases did not have an estimated

fetal weight closer to delivery because the biparietal

diameter measurement was not available (presenting part

too low in the maternal pelvis).

The mean estimated fetal weight customised centiles of

cases with and without IUGR were 16.4 vs 41.7 for low

skinfold thickness, 22.8 vs 42.7 for ponderal index <25th

centile and 26.3 vs 40.5 for mid-arm circumference to

occipito-frontal circumference <�1 SD; these differences

are statistically significant ( P < 0.001, independent two-

sided t test) for low skinfold thickness and low ponderal

index only ( P > 0.05 for mid-arm circumference to

occipito-frontal circumference ratio). The test perform-

ances of the 5th estimated fetal weight customised centile

and the 10th estimated fetal weight customised centile are

presented in Tables 1 and 2. The principal results of interest

are the likelihood ratios for a positive test (likelihood

ratioþ) as these indicate the value of a positive test in

identifying IUGR; both the 5th and 10th estimated fetal

weight customised centiles cutoffs generate either low or

moderately useful likelihood ratios only14.

Fetal abdominal area velocity and at least one anthro-

pometric measurement was available for 257 of the 274

cases. Cutoff Z scores of �2, �1.55 and �1.5 were chosen

for the prediction of low skinfold thickness, low ponderal

index and mid-arm circumference to occipito-frontal cir-

cumference ratio <�1 SD, respectively; likelihood ratios

for positive tests were 10.4 (95% CI 3.9–26), 9.5 (95% CI

4.6–19) and 4.7 (95% CI 2.3–8.4), respectively7. When the

positive likelihood ratios for fetal abdominal area velocity

Table 2. Test performance of customised estimated fetal weight centiles in the prediction of three measures of neonatal nutritional status.

Test Outcome Pre-test

probability

(%)

LRþ(95% CI)

LR�(95% CI)

Post-test

probabilityþ(%)

Change in

probabilityþ(%)

Post-test

probability�(%)

Change in

probability�(%)

Customised estimated

fetal weight

<5th centile

Skinfold thickness

<10th centile

13.3 4.9 (2.7–6.3) 0.84 (0.6– 0.96) 42.9 29.7 11.3 1.9

Customised estimated

fetal weight

<5th centile

Ponderal index

<25th centile

14.7 6.8 (4.5–10.6) 0.84 (0.64–0.96) 53.8 39.1 12.6 2.1

Customised estimated

fetal weight

<5th centile

Mid-arm to

occipito-frontal

circumference ratio

<�1 SD

7.4 6.3 (3.7– 14) 0.79 (0.47–0.97) 33.3 25.9 5.9 1.5

Customised estimated

fetal weight

<10th centile

Skinfold thickness

<10th centile

13.3 4.5 (2.6–8.6) 0.7 (0.58–0.77) 38 24.7 8.4 4.9

Customised estimated

fetal weight

<10th centile

Ponderal index

<25th centile

14.7 4.1 (2.5–7.2) 0.65 (0.55–0.74) 41.4 26.7 10.1 4.6

Customised estimated

fetal weight

<10th centile

Mid-arm to

occipito-frontal

circumference ratio

<�1 SD

7.4 7.1 (3.5–24.1) 0.76 (0.66–0.83) 27.3 19.9 3.8 3.6

LRþ ¼ Likelihood ratio for a positive test. LR� ¼ Likelihood ratio for a negative test.

PREDICTION OF IUGR 413

D RCOG 2003 Br J Obstet Gynaecol 110, pp. 411–415

are compared with those of the fifth estimated fetal weight

customised centile, there is no observable significant dif-

ference in the prediction of any of the three outcomes

( P ¼ 0.22, P ¼ 0.42 and P ¼ 0.2, for ponderal index, mid-

arm circumference to occipito-frontal circumference ratio

and skinfold thickness, respectively). When compared with

the 10th estimated fetal weight customised centile, the fetal

abdominal area velocity positive likelihood ratio is sig-

nificantly better in the prediction of a low ponderal index

( P ¼ 0.04) but there is no significant difference in the

prediction of either low skinfold thickness or mid-arm

circumference to occipito-frontal circumference ratio

( P ¼ 0.28, P ¼ 0.71, respectively)14.

DISCUSSION

The ability to identify the genuinely growth restricted

infant in the late third trimester with standard fetal biom-

etry and easily obtained maternal data would represent a

widely available and potentially important advance in

antenatal care. We have recently demonstrated a moder-

ately strong relationship between customised birthweight

centiles and anthropometric features of IUGR in the same

study population15 and so it is reasonable to hypothesise

that customised estimated fetal weight centiles might use-

fully predict the delivery of a growth restricted infant.

There continues to be a need to be able to accurately

identify the growth restricted infant prior to delivery in

order to begin to be able to reduce the incidence of

antepartum fetal loss by instituting closer monitoring and

expediting delivery. Adjusting fetal measurements to take

account of fetal and maternal physiological variables

improves test performance when compared with a single

estimate of fetal size3 but our results indicate that custom-

ised fetal weight standards at best perform only moderately

well in the prediction of any of the three anthropometric

measures of fetal growth achievement.

Single estimates of fetal size, amniotic fluid volume and

umbilical artery resistance are poor predictors of IUGR3,4,16,

whereas growth velocity of the fetal abdominal area is use-

ful, at least in the research setting7. We have compared the

performance of customised fetal weight centiles with growth

velocity of the fetal abdominal area and found the latter to

have higher likelihood ratios for a positive test although this

was only statistically significant for one comparison in the

same low risk population.

Estimating fetal growth velocity rather than size appears

to offer most promise but sequential measurements inev-

itably attract measurement errors and reduce repeatability. It

remains to be seen whether calculating growth velocity

retains its apparent utility in the prediction of IUGR when

it is performed in a routine clinical setting. Customised

estimated fetal weight centiles are less likely to be adversely

affected by issues of measurement error and repeatability as

only one estimate of fetal size is required. Novel methods of

fetal imaging might improve our ability to identify IUGR in

the future but this is not certain and the techniques are not

immediately available or widely applicable17,18.

Therefore, at present, the use of a simple test such as

serial biometry would appear to be an area worthy of

further investigation in large clinical trials.

Acknowledgements

Dr P. Owen thanks Wellbeing, the charitable arm of the

RCOG for financial support. There is no conflict of interest.

References

1. Fay RA, Dey PL, Saadie CM, et al. Ponderal index; a better definition

of the at risk group with intrauterine growth problems than birth-

weight for gestational age in term infants. Aust N Z J Obstet Gynaecol

1991;31:17 –19.

2. Walther FJ, Ramaekers LHJ. Neonatal morbidity of SGA infants in

relation to their nutritional status at birth. Acta Paediatr Scand

1982;71:437 –440.

3. Owen P, Burton K, Ogston S, Khan KS, Howie P. Unconditional and

conditional standards in the prediction of true intrauterine growth

restriction. Ultrasound Obstet Gynecol 2000;16:439 –444.

4. Chang TC, Robson SC, Spencer JAD, Gallivan S. Identification of

fetal growth retardation; Comparison of Doppler waveform indices

and serial ultrasound measurements of abdominal circumference and

fetal weight. Obstet Gynecol 1993;82:230 –236.

5. Mongelli M, Sverker E, Tambyrajia R. Screening for fetal growth

restriction: a mathematical model of the effect of time interval and

ultrasound error. Obstet Gynecol 1998;92:908–912.

6. Owen P, Maharaj S, Khan KS, Howie P. Time interval influences

performance of growth velocity in predicting fetal growth restriction.

Obstet Gynecol 2001;97:499– 504.

7. Owen P, Khan KS. Fetal growth velocity in the prediction of intra-

uterine growth retardation in a low risk population. Br J Obstet

Gynaecol 1998;105:536 –540.

8. Sanderson DA, Wilcox MA, Johnson IR. The individualised birth-

weight ratio: a new method of identifying intrauterine growth retar-

dation. Br J Obstet Gynaecol 1994;101:310– 314.

9. de Jong CLD, Gardosi J, Dekker GA, Colenbrander GJ, van

Geijn HP. Application of a customised birthweight standard in the

assessment of perinatal outcome in a high risk population. Br J Obstet

Gynaecol 1997;104:531 –535.

10. Owen P, Donnet L, Ogston S, Christie AD, Patel N, Howie PW.

Standards for ultrasound fetal growth velocity. Br J Obstet Gynaecol

1996;103:60– 69.

11. Chien PF, Owen P, Khan KS. Validity of ultrasound estimation of

fetal weight. Obstet Gynecol 2000;95:856 –860.

12. Gardosi JO, Mongelli JM, Wilcox M, Chang A. An adjustable fetal

weight standard. Ultrasound Obstet Gynecol 1995;6:168 –174.

13. Jaeschke R, Guyatt G, Sackett DL. For the evidence-based medicine

working group. Users’ guide to the medical literature. III: How to use

an article about a diagnostic test (2 parts). JAMA 1994;271:389– 391,

703– 707.

14. Efron B, Tibshirami RJ. An introduction to the Bootstrap. Chapter 11:

The jacknife. Monographs on Statistics and Applied Probability 57.

Boca Raton (FL): Chapman & Hall, 1993.

15. Owen P, Farrell T, Hardwick JCR, Khan KS. Relationship between

414 P. OWEN ET AL.

D RCOG 2003 Br J Obstet Gynaecol 110, pp. 411–415

customised birthweight centiles and neonatal anthropometric features

of growth restriction. Br J Obstet Gynaecol 2002;109:658 –662.

16. Owen P, Khan KS, Howie P. Single and serial estimates of amniotic

fluid volume and umbilical artery resistance in the prediction of intra-

uterine growth restriction. Ultrasound Obstet Gynecol 1999;13(6):

415–419.

17. Baker PN, Johnson IR, Gowland PA, et al. Measurement of fetal liver,

brain and placental volumes with echo-planar magnetic resonance

imaging. Br J Obstet Gynaecol 1995;102:35–39.

18. Chang FM, Liang RI, Ko HC, Tao BL, Chang CH, Yu CH. Three

dimensional ultrasound assessed fetal thigh volumetry in predicting

birthweight. Obstet Gynecol 1997;90:331 –339.

Accepted 21 January 2003

PREDICTION OF IUGR 415

D RCOG 2003 Br J Obstet Gynaecol 110, pp. 411–415