Customizing fetal biometric charts - Wiley Online Library

Ultrasound Obstet Gynecol 2003; 22: 271–276 Published online 1 August 2003 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/uog.196

Customizing fetal biometric charts M. W. PANG*, T. N. LEUNG†, D. S. SAHOTA†, T. K. LAU† and A. M. Z. CHANG† Department of Obstetrics and Gynaecology, Prince of Wales Hospital and †The Chinese University of Hong Kong, Hong Kong, SAR

K E Y W O R D S: customization; longitudinal; multilevel modeling; ultrasound fetal biometry

ABSTRACT Objective To study the effects of maternal and pregnancy characteristics on fetal biometric size using longitudinal ultrasound measurements and to construct customized models for fetal biometric size charts. Methods A cohort of 533 healthy pregnant women with normal singleton pregnancies were recruited for regular ultrasound examination for fetal biometry between 24 and 40 weeks’ gestation. Multilevel modeling was used to construct models of fetal head size, femur length and abdominal circumference. Variables of maternal and pregnancy characteristics including booking weight and height, age, parity and fetal sex were included in the construction of the customized fetal biometric size charts. Results Increased fetal head size and abdominal circumference were significantly associated with extremes of maternal age. Maternal height had a statistically significant influence on biparietal diameter. Maternal booking weight had an influence on fetal abdominal circumference and femur length. Fetal sex was found to have a statistically significant influence on the final regression models of biparietal diameter, head circumference and femur length. Parity had an influence on fetal head circumference and abdominal circumference. Conclusions Maternal and pregnancy characteristics have a significant influence on in-utero fetal biometry. We produced models to construct customized fetal biometric size charts. Further validation studies are necessary to evaluate the clinical usefulness of such customized fetal biometric size charts. Copyright  2003 ISUOG. Published by John Wiley & Sons, Ltd.

INTRODUCTION Fetal growth assessment has always been a subject of interest in obstetric practice as fetal growth aberration is

related to adverse perinatal outcome1 . To date, antenatal ultrasound assessment of fetal parameters is still the mainstay of growth surveillance. Traditionally, charts of normal fetal biometry have been determined for local populations. Deviations from normal values were defined by statistical cut-off values2 . As neonatal size was found to vary with the characteristics of the population these population-based fetal nomograms needed to be revised regularly to enable their meaningful clinical application3 – 5 . Keeping et al.6 first introduced the concept of customized standard birth-weight charts based on their observations that neonatal birth weights differed among mothers with different pregnancy characteristics. In-utero fetal growth studies also suggested that certain maternal and pregnancy characteristics affected fetal growth7,8 . Fetal growth was found to be different in subgroups of pregnancies with different maternal height, maternal weight, smoking status, ethnic origin, parity and fetal sex8 . Gardosi et al.9 , based on this concept, performed mathematical modeling in which they took into account the effects of pregnancy characteristics to produce a customized birth-weight standard. Subsequent validation studies found that customized birth-weight standards were useful in detecting babies with abnormal body distribution and adverse perinatal outcome10 – 15 . Birth weight was regarded as an outcome measure of fetal growth. For the purposes of clinical application, antenatal detection of fetal growth aberration appeared to be useful in allowing early implementation of intensive surveillance or intervention. It would be invaluable to similarly incorporate the determinants of fetal size into the modeling of fetal biometric charts i.e. to produce customized fetal biometric charts. The potential benefits of customized fetal biometric charts include better distinction between pathological growth aberrations and physiological extremes. In addition, such charts negate the need for frequent revision

Correspondence to: Dr M. W. Pang, Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, Hong Kong SAR (e-mail: [email protected]) Accepted: 26 March 2003

Copyright  2003 ISUOG. Published by John Wiley & Sons, Ltd.

ORIGINAL PAPER

Pang et al.

272

of the normal charts due to changes in the composition, and thus the pregnancy characteristics, of a population. The aims of our study were therefore to test whether fetal biometric size as measured on ultrasound examination is related to maternal and pregnancy characteristics, and to construct customized models for fetal biometric size charts using longitudinal ultrasound measurements of fetal biometric parameters in a cohort of healthy women with normal singleton pregnancies.

METHODS Study population The study was conducted in a university teaching hospital in Hong Kong during the period 1998 to 2000. The delivery rate during the period of the study was approximately 6000 per year. The parturients consist of both high- and low-risk cases and over 98% are ethnically Chinese. Subjects were invited to participate in the study when they attended our clinic for a routine firsttrimester dating ultrasound examination. The inclusion criteria were: (1) both parents ethnically Chinese; (2) good maternal health; (3) singleton pregnancy; (4) certain date of last menstrual period and regular 28–32-day menstrual cycle; (5) gestational ages calculated by last menstrual period and by measurement of crown–rump length or biparietal diameter in the first trimester within 4 days of each other. Pregnancies that were subsequently found to be complicated by congenital abnormalities, gestational diabetes, hypertensive disorders in pregnancy and preterm deliveries were excluded from the construction of the customized normal charts. Written consent was obtained from the women and the study protocol was approved by the institutional review board.

Ultrasound scans Women who fulfilled the inclusion criteria and who consented to participate were assigned an appointment from 24 weeks’ gestation onwards. Serial ultrasound fetal biometric measurements, including biparietal diameter (outer–outer) (BPDoo), biparietal diameter (outer–inner) (BPDoi), head circumference (HC), abdominal circumference (AC) and femur length (FL), were measured at 2–4-week intervals until delivery. The ultrasound examinations were specifically arranged for the construction of the fetal size charts. All examinations were performed by one of two experienced operators using an Apogee 800PLUS ultrasound machine (ATL, A Philips Medical System Company, Andover, MA, USA) with a 3.5-MHz curvilinear probe. Fetal head measurements were made in the axial plane at the level where the continuous midline echo is broken by the cavum septi pellucidi in the anterior third16 . Measurements of the BPDoi were made from the leading edge of the echo from the proximal fetal skull to the leading edge


of the echo from the distal fetal skull. Measurements of the BPDoo were made in a similar way only the leading edge of the echo from the distal fetal skull was taken. The HC was measured by placing elliptical calipers over the four points of the BPDoo and the occipitofrontal diameter, the latter being measured in the same plane between the leading edge of the frontal bone and the outer border of the occiput16 . The AC was measured on a transverse section through the fetal abdomen as described by Campbell and Wilkin1 . The femur was identified and the transducer rotated until the full femoral diaphysis was seen in a plane as close as possible to 90◦ to the ultrasound beam. A straight measurement from one end of the diaphysis to the other end was made2 . Before commencement of the study, intra- and interobserver variations in the measurements of the two operators were estimated17,18 . The intraobserver repeatability coefficients for Observers 1 and 2, respectively, were: ± 0.88 in both cases for BPDoi, 0 and ± 0.88 for BPDoo, ± 4.10 and ± 1.23 for HC, ± 4.38 and 1.52 for AC and ± 0.88 for FL. The interobserver limits of agreement were − 1.4 to 1.4 for BPDoi, − 1.95 to 1.95 for BPDoo, − 1.99 to 1.94 for HC, − 1.12 to 0.98 for AC and − 0.75 to 0.68 for FL.

Customization The independent variables tested were selected based on reported regression models of fetal biometric size charts and clinical variables which have been shown to have a significant impact on the final birth weight of a pregnancy2,16,19 – 22 . They included the following variables: (1) gestational age and its polynomials (calculated as departures from the mean of the study population); (2) maternal characteristics: (2a) maternal height and its polynomials; (2b) maternal weight and its polynomials; (2c) maternal age (≤ 20, 20–35, ≥ 35); (3) pregnancy characteristics: (3a) parity (para 0, para 1, para 2+); (3b) fetal sex (female = 0, male = 1).

Statistical analysis The Statistical Package for Social Sciences for Windows version 10.0 (SPSS Inc, Chicago, IL, USA) was used to test for the colinearity of the independent variables. MLwiN Version 1.123 (Multilevel Models Project, Institute of Education, London, UK) was used for the construction of the models for the fetal biometric size charts. A two-level model was used to develop the fetal biometric size charts. The level 1 variance was defined as the variation between measurements from the same pregnancy at different gestational ages. The level 2 variance was defined as the inter-pregnancy variation. Stepwise regression was performed. Only independent variables with a significant improvement in the goodnessof-fit to the model as assessed by the deviance statistics (χ2 with P ≤ 0.05) were included in the final regression models23 .

Ultrasound Obstet Gynecol 2003; 22: 271–276.

Customizing fetal biometric charts

273

The total variance of the model was calculated using MLwiN Version 1.123 . The SD was calculated as the square root of the total variance. The 10th and 90th centiles were derived as ± 1.28 SD from the regression mean.

RESULTS A total of 533 women were recruited for the study. Thirty-three of them were subsequently found to have pregnancies complicated with gestational diabetes or hypertensive disorders of pregnancy and these were excluded from the construction of the fetal biometric size charts. Therefore 500 women with 2349 serial ultrasound examinations were included in the construction of the charts. The number of ultrasound assessments for each subject ranged from 3 to 9. The BPDoo, BPDoi, HC, AC and FL were normally distributed with gestational age. Table 1 shows the number of ultrasound examinations performed per completed gestational week in the study population. Colinearity statistics showed no correlation between the independent variables tested (r ≤ 0.8)24 . The mean gestational age of the study population was 32.07 weeks. The mean (SD) maternal height was 157.5 (5.4) cm and mean maternal booking weight was 52.8 (7.6) kg. The regression models for BPDoo, BPDoi, HC, AC and FL and their respective variances are given in Tables 2–6. Of the maternal characteristics, maternal height had a statistically significant influence on BPDoo, maternal booking weight had a significant influence on fetal AC and FL and maternal age influenced all fetal parameters except FL. Of the pregnancy characteristics, fetal sex had a statistically significant influence on BPDoo, BPDoi, HC and FL and parity had an influence on fetal HC and AC. Figures 1–3 give examples of the expected fetal biometric size measurements in two hypothetical pregnancies Table 1 Number of ultrasound examinations performed per completed gestational week Completed gestational weeks 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Total

Number of measurements 147 140 138 144 134 150 149 140 120 151 153 140 154 235 145 86 23 2349


Table 2 Customized outer–outer biparietal diameter (BPDoo) regression model Parameter

BPDoo (mm)

Constant Gestation (GT) Gestation2 (GT2 ) Sex Male Female Maternal height Maternal age ≤ 20 years

Adjustment

P

90.40 + 2.36 mm/week ≤ 0.05 − 0.12 mm/week ≤ 0.05 + 1.21 mm + 0 mm + 0.04 mm/cm + 2.11 mm

≤ 0.05 ≤ 0.05 ≤ 0.05

Variance = 6.973 + 0.33 GT − 0.0724 GT2 . Maternal height and weight adjusted from population means of 157.5 cm and 52.8 kg; gestational age adjusted from population mean of 32.07 weeks.

Table 3 Customized outer–inner biparietal diameter (BPDoi) regression model Parameter

BPDoi (mm)

Constant Gestation (GT) Gestation2 (GT2 ) Sex Male Female Maternal age ≤ 20 years

Adjustment

P

88.46 + 2.32 mm/week ≤ 0.05 − 1.18 mm/week ≤ 0.05 + 1.20 mm + 0 mm + 1.87 mm

≤ 0.05 ≤ 0.05


Table 4 Customized head circumference (HC) regression model Parameter

HC (mm)

Constant Gestation (GT) Gestation2 (GT2 ) Sex Male Female Parity Parity 1 Maternal age 20–35 years

Adjustment

P

301.10 + 0.74 mm/week ≤ 0.05 − 0.03 mm/week ≤ 0.05 + 0.33 mm + 0 mm

≤ 0.05

− 0.16 mm − 0.18 mm

≤ 0.05 ≤ 0.05

Variance = 0.733 + 0.038 GT − 0.006 GT2 − (0.28 + 0.006 GT − 0.002 GT2 ) Parity 1. Maternal height and weight adjusted from population means of 157.5 cm and 52.8 kg; gestational age adjusted from population mean of 32.07 weeks.

with distinct maternal and pregnancy characteristics; the 50th centiles and their respective 95% limits of agreement have been plotted against gestational age.

DISCUSSION To our knowledge, this is the first study to use multilevel modeling in the construction of customized fetal biometric


Pang et al.

274 Table 5 Customized abdominal circumference (AC) regression model AC (mm)

Constant Gestation (GT) Gestation2 (GT2 ) Maternal weight Parity Parity 0 Parity 1 Maternal age ≥ 35 years

Adjustment

P

294.52 + 1.03 mm/week − 0.01 mm/week2 + 0.01 mm/kg

≤ 0.05 ≤ 0.05 ≤ 0.05

− 0.29 mm − 0.16 mm + 0.36 mm

≤ 0.05 ≤ 0.05 ≤ 0.05

2

Abdominal circumference (mm)

Parameter

380

Table 6 Customized femur length (FL) regression model

Constant Gestation (GT) Gestation2 (GT2 ) Gestation3 (GT3 ) Maternal weight Sex Male Female

FL (mm)

230

Adjustment

P

+ 2.20 mm/week − 0.06 mm/week2 − 0.001 mm/week3 + 0.03 mm/kg

≤ 0.05 ≤ 0.05 ≤ 0.05 ≤ 0.05

− 0.43 mm + 0 mm

≤ 0.05

66.76

180

Figure 2 Comparison of expected abdominal circumference measurements in two hypothetical pregnancies after customization of distinct maternal and pregnancy characteristics. The 50th centiles and their 95% limits of agreement are shown. Large (95% limits of agreement, dashed lines; 50th centile, solid line between them): parity ≥ 2, maternal booking weight at 90th centile of the population, maternal age ≥ 35 years; small (95% limits of agreement, dotted lines; 50th centile, solid line between them): parity 0, maternal booking weight at 10th centile of the population, maternal age < 35 years.

85 80 75 Femur length (mm)

350 330

70 65 60

310

55

290

50

270

45

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Gestation (weeks)

250 230 210

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Gestation (weeks)


Head circumference (mm)

280

3

Variance = 1.233 + 0.122 GT − 0.0019 GT + 0.002 GT + 0.00078 GT4 + (0.56 − 0.072 GT − 0.006 GT2 ) Parity 1. Maternal height and weight adjusted from population means of 157.5 cm and 52.8 kg; gestational age adjusted from population mean of 32.07 weeks.

Parameter

330

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 Gestation (weeks)

Figure 1 Comparison of expected head circumference measurements in two hypothetical pregnancies after customization of distinct maternal and pregnancy characteristics. The 50th centiles and their 95% limits of agreement are shown. Large (95% limits of agreement, dashed lines; 50th centile, solid line between them): male fetus, parity 0, maternal age ≤ 20 years; small (95% limits of agreement, dotted lines; 50th centile, solid line between them): female fetus, parity 1, maternal age between 20 and 35 years.


Figure 3 Comparison of expected femur length measurements in two hypothetical pregnancies after customization of distinct maternal and pregnancy characteristics. The 50th centiles and their 95% limits of agreement are shown. Large (95% limits of agreement, dashed lines; 50th centile, solid line between them): female fetus, maternal booking weight at 90th centile of the population; small (95% limits of agreement, dotted lines; 50th centile, solid line between them): male fetus, maternal weight at 10th centile of the population.

size charts. Conventionally, fetal biometric size charts were constructed with ultrasound measurement of each fetus only once18 . The use of longitudinal data creates a


Customizing fetal biometric charts number of statistical problems8,19 . The most important drawback is the underestimation of the variability between the subjects as repeated measures are taken in the same individual8 . Some investigators tried to overcome this problem by generating an individual growth curve for each subject. The between-subject means of the estimated coefficients are taken as the basis of an average curve20 . Using this approach, the statistical estimates can be unreliable. Multilevel modeling offers several advantages over conventional statistics for the processing of longitudinal growth data. First, the interval between each measurement at the individual level need not be fixed8 . This allows greater flexibility in the process of data collection. It also gives better precision in the assessment of the relationship between the independent and dependent variables as there is no need to aggregate the time factor for analysis. Second, it allows modeling of variability (intra- and inter-subject) at different levels8 . Third, it provides a framework for the calculation of reference intervals for both fetal size and fetal growth using the same set of longitudinal data19 . Fourth, covariates such as maternal and pregnancy characteristics can be added into the model allowing development of a customized model for fetal size in individual pregnancies8 . Unlike other reported studies, we decided to construct customized models of individual fetal biometric parameters rather than fetal weight20 . This approach eliminates the potential problem of erroneous estimation of fetal weight using formulae for its calculation. In addition, it allows direct assessment of the characteristics of the growth aberration by comparison of the deviation of the individual fetal biometric parameters from the expected values i.e. symmetrical vs. asymmetrical growth restriction. Mongelli and Gardosi20 reported that pregnancy characteristics such as ethnicity, maternal booking height, maternal booking weight, parity and smoking had a significant impact on the in-utero estimated fetal weight. Smulian et al.21 detected a statistically significant yet small difference in individual fetal biometric measurements in male and female fetuses. The effect of some of these variables on fetal biometry was confirmed in our study. It is of interest to note that in our study, in addition to the determinants as described by other investigators, we were able to demonstrate that the extremes of maternal age give rise to a statistically significant increase in fetal HC and AC. This observation is independent of other covariates as examined in colinearity statistics. Our results have shown that the effect of maternal and pregnancy characteristics on fetal biometry can be observed as early as 24 weeks’ gestation. Due to the design of this study, we are unable to comment on whether such differences exist at earlier gestational ages. If a similar significant effect could be demonstrated at 16–20 weeks of gestation, there would be important implications on the use of fetal biometry for dating purposes. Our study has also shown that certain maternal and pregnancy characteristics have only a minimal effect on FL, but


275

they have a much larger effect on HC and AC. This may imply that the growth of different fetal parts and organs is influenced by different regulatory pathways. In conclusion, the results from our study confirm the observation that maternal and pregnancy characteristics affect fetal biometric measurements. These findings are at least of physiological interest. We have produced models to construct customized fetal biometric size charts to allow for these effects. Further clinical studies are necessary to evaluate the performance of such customized fetal biometric charts compared with unmodified populationbased biometric charts in differentiating fetuses with growth aberrations from those whose growth is at the physiological extremes.

ACKNOWLEDGMENTS We acknowledge the assistance of C. F. Poon, S. M. Wong, P. Y. Chan for their support in data collection and data entry during the research period.

REFERENCES 1. Campbell S, Wilkin D. Ultrasonic measurement of fetal abdominal circumference in the estimation of fetal weight. Br J Obstet Gynaecol 1975; 82: 689–697. 2. Chitty LS, Altman DG, Henderson A, Campbell S. Charts of fetal size: 4. Femur length. Br J Obstet Gynaecol 1994; 101: 132–135. 3. Wohlfahrt J, Melbye M, Christens P, Anderson AM, Hjalgrim H. Secular and seasonal variation of length and weight at birth. Lancet 1998; 352: 1990–1993. 4. Li DK, Ni HY, Schwartz SM, Daling JR. Secular change in birthweight among southeast Asian immigrants to the United States. Am J Public Health 1990; 80: 685–688. 5. Rosenberg M. Birth weights in three Norwegian cities, 1860–1984. Secular trends and influencing factors. Ann Hum Biol 1988; 15: 275–288. 6. Keeping JD, Chang A, Morrison J. Birth weight: Analysis of variance and the linear additive model. Br J Obstet Gynaecol 1979; 86: 437–442. 7. Deter RL, Harrist RB, Hadlock FP, Pointdexter AN. Longitudinal studies of fetal growth with the use of dynamic image ultrasonography. Am J Obstet Gynecol 1982; 143: 545–554. 8. Rice N, Leyland A. Multilevel models: applications to health data. J Health Serv Res Policy 1996; 1: 154–164. 9. Gardosi J, Mongelli M, Wilcox M, Chang A. An adjustable fetal weight standard. Ultrasound Obstet Gynecol 1995; 6: 168–174. 10. Mongelli M, Gardosi J. Reduction of false-positive diagnosis of fetal growth restriction by application of customized fetal growth standards. Obstet Gynecol 1996; 88: 844–848. 11. Vik T, Markestad T, Ahlsten G, Gebre-Medhin M, Jacobsen G, Hoffman HJ. Body proportions and early neonatal morbidity in small-for-gestational age infants of successive births. Acta Obstet Gynecol Scand Suppl 1997; 165: 76–81. 12. Sanderson DA, Wilcox MA, Johnson IR. Relative macrosomia identified by the individualised birthweight ratio (IBR). Acta Obstet Gynecol Scand 1994; 73: 246–249. 13. 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. 14. Sanderson DA, Wilcox MA, Johnson IR. The individualised birthweight ratio: a new method of identifying intrauterine growth retardation. Br J Obstet Gynaecol 1994; 101: 310–314.


276 15. Clausson B, Gardosi J, Francis A, Cnattingius S. Perinatal outcome in SGA births defined by customised versus population-based birthweight standards. BJOG 2001; 108: 830–834. 16. Chitty LS, Altman DG, Henderson A, Campbell SC. Charts of fetal size: 2. Head measurements. Br J Obstet Gynaecol 1994; 101: 35–43. 17. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8: 135–160. 18. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 8: 307–310. 19. Royston P. Calculation of unconditional and conditional reference intervals for foetal size and growth from longitudinal measurements. Stat Med 1995; 14: 1417–1436.


Pang et al. 20. Mongelli M, Gardosi J. Longitudinal study of fetal growth in subgroups of a low-risk population. Ultrasound Obstet Gynecol 1995; 6: 340–344. 21. Smulian JC, Campbell WA, Rodis JF, Feeney LD, Fabbri EL, Vintzileos AM. Gender-specific second-trimester biometry. Am J Obstet Gynecol 1995; 173: 1195–1201. 22. Chitty LS, Altman DG, Henderson A, Campbell S. Charts of fetal size: 3. Abdominal measurements. Br J Obstet Gynaecol 1994; 101: 125–131. 23. Ian Plewis. Reading progress. In A guide to MLn for New Users, Woodhouse G (ed). Institute of Education, University of London: London, 1995; 115–119. 24. Katz MH. Relationship of independent variables to one another. In Multivariable analysis. A practical guide for clinicians (1st edn). Cambridge University Press: Cambridge; 1999; 56.
